DYNA MIC

SYSTEMS
Modeling, Simulation, and Analysis

BINGEN YANG
INNA ABRAMOVA
Dynamic Systems

Presenting students with a comprehensive and efficient approach to the modeling, simula-
tion, and analysis of dynamic systems, this textbook addresses mechanical, electrical, ther-
mal and fluid systems, feedback control systems, and their combinations. It features a robust
introduction to fundamental mathematical prerequisites, suitable for students from a range
of backgrounds; clearly established three-key procedures – fundamental principles, basic
elements, and ways of analysis – for students to build on in confidence as they explore new
topics; over 300 end-of-chapter problems, with solutions available for instructors, to solidify
a hands-on understanding; and clear and uncomplicated examples using MATLAB®/
Simulink®, and Mathematica®, to introduce students to computational approaches. With
a capstone chapter focused on the application of these techniques to real-world engineering
problems, this is an ideal resource for a single-semester course in dynamic systems for
students in mechanical, aerospace, and civil engineering.

Bingen “Ben” Yang is Professor of Aerospace and Mechanical Engineering at the University
of Southern California, where he has taught for more than three decades. Being an active
researcher, he has more than 230 publications in the areas of structures, dynamics, vibra-
tions, controls, and mechanics. His current research interest lies in modeling, analysis,
control and design of flexible structures, electromechanical systems, and computational
methods for problems in engineering applications. A fellow of the American Society of
Mechanical Engineers, Dr. Yang received his Ph.D. in mechanical engineering from the
University of California at Berkeley.

Inna Abramova is a senior lecturer of Aerospace and Mechanical Engineering at the

University of Southern California, where she has taught for seven years. Her research
interests lie in the areas of dynamics and controls, with emphasis on collaborative operations
of autonomous agents. Dr. Abramova received her Ph.D. in mechanical engineering from
the University of Southern California.
Dynamic Systems
Modelling, Simulation, and Analysis
Bingen Yang
University of Southern California

Inna Abramova
University of Southern California
University Printing House, Cambridge CB2 8BS, United Kingdom
One Liberty Plaza, 20th Floor, New York, NY 10006, USA
477 Williamstown Road, Port Melbourne, VIC 3207, Australia
314–321, 3rd Floor, Plot 3, Splendor Forum, Jasola District Centre,
New Delhi – 110025, India
103 Penang Road, #05–06/07, Visioncrest Commercial, Singapore 238467

Cambridge University Press is part of the University of Cambridge.

It furthers the University’s mission by disseminating knowledge in the pursuit of
education, learning, and research at the highest international levels of excellence.

www.cambridge.org
Information on this title: www.cambridge.org/highereducation/isbn/9781107179790
DOI: 10.1017/9781316841051
© Bingen Yang and Inna Abramova 2022
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2022
Printed in the United Kingdom by TJ Books Limited, Padstow Cornwall
A catalogue record for this publication is available from the British Library.
Library of Congress Cataloging-in-Publication Data
Names: Yang, Bingen, author. | Abramova, Inna, author.
Title: Dynamic systems : modelling, simulation, and analysis / Bingen Yang, University of Southern California,
Inna Abramova, University of Southern California.
Description: Cambridge, United Kingdom ; New York, NY : Cambridge University Press, 2022. | Includes
bibliographical references and index.
Identifiers: LCCN 2021044739 (print) | LCCN 2021044740 (ebook) | ISBN 9781107179790 (hardback) | ISBN
9781316841051 (ebook)
Subjects: LCSH: Dynamics. | BISAC: TECHNOLOGY & ENGINEERING / General
Classification: LCC TA352 .Y36 2022 (print) | LCC TA352 (ebook) | DDC 620.1/04–dc23
LC record available at https://lccn.loc.gov/2021044739
LC ebook record available at https://lccn.loc.gov/2021044740
ISBN 978-1-107-17979-0 Hardback
Cambridge University Press has no responsibility for the persistence or accuracy of
URLs for external or third-party internet websites referred to in this publication
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.
Contents

Preface page ix
Acknowledgments xiii

1 Introduction 1
1.1 General Concepts 1
1.2 Classification of Dynamic Systems 4
1.3 System Model Representations 6
1.4 Analogous Relations 11
1.5 Software for Computation 13
1.6 Units 15

2 Fundamentals of Mathematics 16
2.1 Vector Algebra 16
2.2 Matrix Algebra 24
2.3 Complex Numbers 34
2.4 The Laplace Transform 37
2.5 Differential Equations 63
2.6 Solution of Linear Ordinary Differential Equations with Constant
Coefficients 70
2.7 Transfer Functions and Block Diagrams of Time-Invariant Systems 95
2.8 Solution of State Equations via Numerical Integration 98
Chapter Summary 102
References 103
Problems 103

3 Mechanical Systems 114

3.1 Fundamental Principles of Mechanical Systems 115
3.2 Translational Systems 120
3.3 Rotational Systems 144
3.4 Rigid-Body Systems in Plane Motion 169
3.5 Energy Approach 177
3.6 Block Diagrams of Mechanical Systems 181

v
vi Contents

3.7 State-Space Representations 185

3.8 Dynamic Responses via MATLAB 191
3.9 Dynamic Responses via Mathematica 198
Chapter Summary 204
References 204
Problems 204

4 Electrical Systems 222

4.1 Fundamentals of Electrical Systems 223
4.2 Concept of Impedance 230
4.3 Kirchhoff’s Laws 242
4.4 Passive-Circuit Analysis 244
4.5 State-Space Representations and Block Diagrams 258
4.6 Passive Filters 271
4.7 Active-Circuit Analysis 277
4.8 Dynamic Responses via Mathematica 297
Chapter Summary 306
References 306
Problems 307

5 Thermal and Fluid Systems 334

5.1 Fundamental Principles of Thermal Systems 335
5.2 Basic Thermal Elements 339
5.3 Dynamic Modeling of Thermal Systems 346
5.4 Fundamental Principles of Fluid Systems 353
5.5 Basic Elements of Liquid-Level Systems 356
5.6 Dynamic Modeling of Liquid-Level Systems 363
5.7 Fluid Inertance 368
5.8 The Bernoulli Equation 370
5.9 Pneumatic Systems 374
5.10 Dynamic Response via MATLAB and Simulink 378
Chapter Summary 384
References 384
Problems 384

6 Combined Systems and System Modeling Techniques 396

6.1 Introduction to System-Level Modeling 397
6.2 System Modeling Techniques 398
6.3 Fundamentals of Electromechanical Systems 425
 Contents vii

6.4 Direct Current Motors 433

6.5 Block Diagrams in the Time Domain 457
6.6 Modeling and Simulation by Simulink 464
6.7 Modeling and Simulation by Mathematica 475
Chapter Summary 485
References 486
Problems 486

7 System Response Analysis 502

7.1 System Response Analysis in the Time Domain 502
7.2 Stability Analysis 544
7.3 System Response Analysis in the Frequency Domain 558
7.4 Time Response of Linear Time-Varying and Nonlinear Systems 571
Chapter Summary 577
References 578
Problems 578

8 Introduction to Feedback Control Systems 589

8.1 General Concepts 589
8.2 Advantages of Closed-Loop Control Systems 595
8.3 PID Control Algorithm 601
8.4 Control System Analysis 611
8.5 The Root Locus Method 619
8.6 Analysis and Design by the Root Locus Method 632
8.7 Additional Examples of Control Systems 648
Chapter Summary 653
References 654
Problems 654

9 Application Problems 666

9.1 Vibration Analysis of a Car Moving on a Bumpy Road 667
9.2 Speed Control of a Coupled Engine–Propeller System 686
9.3 Modeling and Analysis of a Thermomechanical System (a Bimetallic Strip
Thermometer) 697
9.4 Modeling and Analysis of an Electro-Thermo-Mechanical System (a
Resistive-Heating Element) 702
9.5 Feedback Control of a Liquid-Level System for Water Purification 711
9.6 Sensors, Electroacoustic, and Piezoelectric Devices 723
References 756
viii Contents

Appendices
A. Units and Conversion Table 758
B. A Brief Introduction to MATLAB and Simulink 761
C. A Brief Introduction to Wolfram Mathematica 771
Index 778
Preface

The subject of dynamic systems deals with modeling and analysis of temporal behaviors of
machines, devices, equipment, structures, industrial processes, and beyond. Mathematically,
a dynamic system is described by certain time-dependent functions, which are usually
governed by a set of differential equations. Physically, a typical dynamic system consists of
multiple components or subsystems, which fall in different areas, such as mechanics,
thermodynamics, fluid dynamics, electromagnetism, acoustics, and optics. Unlike those
single-field courses, a dynamic systems course is focused on interactions among the compo-
nents of a system, integration of those components from multiple fields in modeling, and
overall system-level behaviors in analysis, which is known as system dynamics. Naturally,
modeling and analysis of dynamic systems requires a variety of techniques, including transfer
function formulations, block diagrams, and state-space representations. Additionally, an in-
depth understanding of system dynamics is crucially important to the development of
feedback control systems, which nowadays are a must in many applications. Therefore,
a dynamic systems course serves as a bridge between single-area studies and system-level
investigations, and it is a preparation for research and development on advanced topics in
science and technology.

Readership
This book introduces a wide range of dynamic systems encountered in engineering. In
particular, mechanical, electrical, thermal and fluid systems, feedback control systems, and
their combinations are modeled, simulated, and analyzed. The material is designed for
a course at undergraduate level, which is typically found in engineering curricula. The text
can also be used by engineers and researchers for self-learning purposes.
The reader is expected to have a background in elementary calculus (differentiation and
integration) and some engineering college courses (such as dynamics, the strength of mater-
ials, electrical circuits, physics, thermodynamics, and fluid dynamics). A knowledge of
differential equations is desirable but not necessary because this topic, along with others, is
fully covered in the text.

ix
x Preface

Special Features
Reflecting the combined 20 plus years of the authors’ experiences in teaching dynamic
systems, the book has the following special features.
First, three important keys in system modeling (namely, fundamental principles, basic
elements, and ways of analysis) are emphasized throughout the text. In the authors’ opinion,
learning this three-key modeling concept can greatly enhance the reader’s skills to develop
mathematical models for dynamic systems in various engineering problems.
Second, the relevant mathematical topics, including differential equations, the Laplace
transform, matrices, complex numbers, and the numerical solution of state equations, are
thoroughly reviewed. Thus, readers of different mathematical backgrounds should be able to
master the contents of the book with ease.
Third, computer simulations of dynamic systems using both MATLAB/Simulink and
Mathematica are illustrated in examples, and also used to solve a variety of application
problems. The coverage of these two powerful software packages, which is not seen in most
dynamic systems texts, provides the reader with computational flexibility in investigation of
system dynamics.
Last, but not least, a cluster of practical problems of multi-component dynamic systems is
presented in the last chapter (Chapter 9). This helps the reader digest the materials that are
presented in the previous chapters, and demonstrates the utility of software packages in
modeling, simulation, and analysis of complicated dynamic systems in engineering.

Organization
This book comprises nine chapters and three appendices.
Chapter 1 introduces the general concepts and classification of dynamic systems in
engineering, and previews commonly used methods and computer software for modeling,
simulation, and analysis of dynamic systems. Moreover, analogous relations among mech-
anical, electrical, thermal, and fluid systems are presented.
Chapter 2 presents the essential mathematical topics that are required for modeling,
simulation, and analysis of dynamic systems. These topics include vectors and matrices,
differential equations, the Laplace transform, complex numbers, transfer functions, block
diagrams, and the solution of state equations by numerical integration.
Chapter 3 covers modeling of mechanical systems, including translational, rotational, and
rigid-body-systems. In the derivation of the equations of motion for a system, three keys are
emphasized: fundamental principles (such as Newton’s laws of motion), basic elements (such
as masses, springs, and dampers), and a way of analysis (for example, free-body diagrams).
Also, transfer function formulations and state-space representations are introduced to
prepare for the analysis and simulation of mechanical systems. In addition, an energy
 Preface xi

approach with Lagrange’s equations is introduced as an alternative method to derive

equations of motion for mechanical systems.
Chapter 4 covers modeling of electrical systems, including passive electrical circuits and
active electrical circuits with isolation amplifiers and operational amplifiers (op-amps). Just
as when dealing with mechanical systems, there are three keys in the modeling of electrical
systems: Kirchhoff’s current and voltage laws as fundamental principles; resistors, inductors,
and capacitors as basic elements; and the loop method and the node method as two ways of
analysis. Also, transfer functions and state equations for electrical systems are presented.
Chapter 5 covers modeling of thermal and fluid systems, in which lumped thermal and
fluid elements are introduced. The concept of free-body diagrams for mechanical systems is
extended to thermal and fluid systems. Furthermore, transfer function formulations, state-
space representations, and block diagrams for thermal and fluid systems are presented.
Chapter 6 deals with dynamic systems consisting of multiple components or subsystems
that are from different fields. Electromechanical systems are mainly considered as demon-
strative examples. To describe the interactions among the components of these systems,
transfer functions, s-domain block diagrams, and state equations are used. Furthermore, to
deal with time-variant and nonlinear dynamic systems, time-domain block diagrams are
introduced, followed by examples of modeling and simulation of combined systems via
Simulink and Mathematica.
Chapter 7 presents the methods for the determination of responses of dynamic systems, in
both the time domain and the frequency domain, and for stability analysis. With the dynamic
models presented in Chapters 3–6, the system response is obtained by either analytical
methods or numerical methods.
Chapter 8 gives a brief introduction to feedback control, in which the basic concepts and
advantages of closed-loop control systems are discussed, and the proportional–integral–
derivative (PID) feedback control algorithm is presented. In addition, the root-locus
method, which is a useful tool for the analysis and design of feedback control systems, is
presented in some detail.
Chapter 9 assembles some relatively complicated problems from engineering applications,
including a vehicle moving on a bumpy road, a coupled engine–propelller system,
a bimetallic strip thermometer, a resistive-heating element, a liquid-level system for water
purification, sensors, piezoelectric devices, and electroacoustic devices. In solution of these
problems, the methods of modeling and analysis that are covered in the previous chapters are
used, and the use of the MATLAB/Simulink and Mathematica software packages in simu-
lation is illustrated.
At the end of Chapters 2–8 there is a summary section, reviewing the main contents and
learning objectives of the chapter, as well as a comprehensive list of problems for each topic.
Solutions for these problems are available at www.cambridge.org/yang-abramova.
Appendix A presents a list of SI units for commonly used physical quantities. Appendices
B and C give brief reviews of the basics of MATLAB/Simulink and Mathematica,
respectively.
xii Preface

Use of the Book in Teaching

The selection of the topics from the book depends on three factors: the instructor’s emphasis,
the student’s background in mathematics, and the course duration (semester or quarter).
For a semester course at the junior level, the coverage of Chapter 1, part of Chapter 2,
Chapters 3 through 7, and part of Chapter 8 is recommended. If time allows, a couple of
sections from Chapter 9 can be considered.
Some courses can omit most of Chapter 2 (Fundamentals of Mathematics) if the overall
mathematical level of the class is adequate. Other courses may skip Chapter 5 (Thermal and
Fluid Systems), in order to address other topics of interest.
With any selected topic, Chapter 2 can always serve as a reference on mathematics.

Instructor’s Solutions Manual

A solutions manual for all the listed exercise problems is prepared by the authors.

Website
The publisher maintains a website for this text at www.cambridge.org/yang-abramova.
Here, students can get access to the MATLAB/Simulink and Mathematica codes developed
for the solutions of the application problems in Chapter 9, a tutorial on usage of MATLAB/
Simulink and Mathematica in modeling and simulation, and other resources relevant to the
book. Also, at www.cambridge.org/yang-abramova, instructors adopting this text can gain
access to the solutions manual.
Acknowledgments

The authors would like to express their deep gratitude to the staff at Cambridge University
Press, especially Steven Elliot, Stefanie Seaton, and Charles Howell, and the copy-editor,
Zoë Lewin, for their assistance and guidance during various stages of the development of this
text.
Many thanks go to the University of Southern California for providing an atmosphere
that encourages teaching innovation and curriculum development.
The reviewers of this book provided insightful comments and valuable suggestions, and
for that the authors are grateful.
The first author thanks his wife, Haiyan, and daughters, Sonia and Tanya, for their
support, patience, and understanding.
The second author thanks her colleagues at the Department of Aerospace and Mechanical
Engineering for their invaluable mentorship and support, and her family for their under-
standing and encouragement all the way.

xiii
1 Introduction

Contents
1.1 General Concepts 1
1.2 Classification of Dynamic Systems 4
1.3 System Model Representations 6
1.4 Analogous Relations 11
1.5 Software for Computation 13
1.6 Units 15

In this introductory chapter, the general concepts and classification of dynamic systems
in engineering are introduced; commonly used methods and computer software for
modeling, simulation, and analysis of dynamic systems are previewed; and the scope of
this book is outlined.

1.1 General Concepts

A dynamic system is a combination of components or subsystems, which, with temporal
characteristics, interact with each other to perform a specified objective. There exists
a variety of dynamic systems in applications, such as machines, devices, appliances, equip-
ment, structures, and industrial processes. Mathematically, a dynamic system is character-
ized by time-dependent functions or variables, which are governed by a set of differential
equations. Physically, the components of a dynamic system may fall in different fields of
science and engineering, such as mechanics, thermodynamics, fluid dynamics, vibrations,
elasticity, electronics, acoustics, optics, and controls.
As an example, an electric motor is a dynamic system consisting of mechanical compo-
nents (like rotating shaft, bearing, and housing), electromagnetic components (such as
magnets, coils, and electrical interconnects), and components for controlling the motor

1
2 1 Introduction

speed (including speed sensor, control logic board, and driver). These components interact
with each other to achieve a desired motor speed. The rotation speed and circuit currents are
time-dependent variables of the motor that are governed by differential equations in the
fields of dynamics and electromagnetism.
Unlike those single-field courses, a dynamic systems course is focused on the interactions
among the components of a dynamic system, the integration of those components from
multiple fields in modeling and solution, and the understanding of overall system response in
analysis, which is known as system dynamics. Thus, in study of dynamic systems, the
knowledge in multiple fields is generally required, and various techniques for modeling
and analysis are naturally adopted.
This book introduces a wide range of dynamic systems encountered in engineering
applications. In particular, mechanical, electrical, thermal, and fluid systems, feedback
control systems, and their combinations are modeled, simulated, and analyzed.

1.1.1 Input and Output

A dynamic system can be graphically illustrated by a diagram. Figure 1.1.1(a) shows
a system with one input and one output, which is known as a single-input single-output
(SISO) system. Here, the word “input” means a cause or excitation and the word
“output” means an effect or response due to the input. For example, for an electric
motor, the applied voltage is an input, and the motor speed is an output. As another
example, for a sailboat, the wind flowing on its sail is an input, and the motion of the
boat in the water is the output. The output of a dynamic system is also called a dynamic
response or a system response.
A dynamic system can have multiple inputs and multiple outputs, as shown in
Figure 1.1.1(b). Such a system is known as a multi-input multi-output (MIMO) system.
The specification of inputs and outputs depends on the problem and the investigator’s
interest. For instance, for an electric motor, the applied voltage and the external load torque,
which is exerted by a machine or device that is driven by the motor, are the inputs to the
motor; the motor speed and the motor power (a product of speed and torque) can be
considered as the outputs. In some applications, like robotics, the motor rotation angle
can also be chosen as an output.

Input Output
System Inputs System Outputs

Figure 1.1.1 Block diagram representation of dynamic systems: (a) a single-input single-output system;
and (b) a multi-input multi-output system
 1.1 General Concepts 3

The dynamic behavior of a system can be characterized by its input–output relations, which
describe the influence of inputs on the system outputs. The input–output relations also
provide a means for interconnection of the components of a system in system-level modeling
and analysis, and for understanding of the dynamic interactions among these components.
Therefore, one major task in the study of dynamic systems is to establish input–output
relations.

1.1.2 Modeling, Simulation, and Analysis

In the advent of computer technologies, model-based design has become a must in the
development of many products and for the operation of industrial processes.
Conventional design often relies on building a prototype system and conducting tests on
it, which can be expensive, time-consuming, and in some cases infeasible. Model-based
design, on the other hand, does not encounter these issues and, more importantly, it can
aid optimal design of complicated dynamic systems, and provide an approach to efficient
and reliable operation of industrial processes.
Model-based design engages in the activities of modeling, simulation, and analysis of
physical systems. These efforts are explained as follows.

Modeling
Modeling is the effort to develop a mathematical model for a given dynamic system, which is
a set of differential equations governing the time-dependent variables and outputs of the
system. There are three important keys in modeling: fundamental principles (like Newton’s
laws), basic elements (such as spring, mass, and damper elements), and ways of analysis (for
instance, free-body diagrams). By a way of analysis, fundamental principles are applied to
basic elements, leading to the governing equations of the system. Throughout this text, these
three keys are emphasized in modeling of mechanical, electrical, thermal, and fluid systems,
as shown in Chapters 3 to 5.

Simulation
Simulation is the effort to determine the dynamic response of a system, which involves
solution of the governing differential equations of the system. For certain simple systems,
analytical solutions are available. In general, system response solutions are obtained by
numerical methods, such as numerical integration algorithms (Section 2.8). In simulation,
the mathematical model of a system is usually converted to one of the following representa-
tions: transfer function formulation, state-space representation, and block diagram repre-
sentation. With these model representations, the dynamic response of a system can be
computed through use of software packages, such as MATLAB/Simulink and Mathematica.
4 1 Introduction

Analysis
Analysis is the effort to examine the physical behaviors of a system, to evaluate the
performance of the system, and to provide guidance for changing the parameters and
configuration of the system in a design process. Analysis is often performed based on system
response solutions obtained by simulation. In this text, three main issues in system response
analysis are addressed: system response analysis in the time domain, stability analysis, and
system response analysis in the frequency domain (see Chapter 7).

1.2 Classification of Dynamic Systems

Dynamic systems can have different types of mathematical models: lumped or distributed;
linear or nonlinear; time-invariant or time-variant; and continuous-time or discrete-time.
The classification of these models depends on the types of governing equations obtained in
system modeling, which is described below.

Distributed Models versus Lumped Models

A distributed model has variables as functions of time and spatial coordinate(s), which are
governed by partial differential equations (PDEs). A distributed model is also called an
infinite-dimensional system because its response is expressed in terms of an infinite number
of coordinates. A lumped model has variables as functions of time only, which are governed
by ordinary differential equations (ODEs). A lumped model is also called a finite-dimensional
system as its response is expressed by a finite number of coordinates. For example, heat
conduction in a continuum is originally described by PDEs. Lumping physical parameters
(such as temperature and heat flow rate) leads to element models (thermal capacitance and
thermal resistance), which eventually yields a lumped model described by ODEs.

Linear Models versus Nonlinear Models

A linear model has variables that are governed by linear differential equations. A nonlinear
model has variables that are governed by nonlinear differential equations. A linear differen-
tial equation about a variable is one which only involves a linear combination of the variable
and its derivatives. A nonlinear differential equation about a variable is one which involves
the products and nonlinear functions of the variable and its derivatives.
For instance, the displacement x of a spring–mass–damper system is modeled by the linear
differential equation

x þ cx_ þ kx ¼ f
m€ (1.2.1)
 1.2 Classification of Dynamic Systems 5

where x_ ¼ dx=dt, m, c, and k are the mass, damping, and spring parameters of the system, and
f is an external force. Equation (1.1.1) represents a linear model. If the system is also subject
to dry friction, its governing equation of motion can be written as

x þ cx_ þ kx þ μN sgnðxÞ
m€ _ ¼f (1.2.2)

where μ is a kinetic friction coefficient, N is a normal force, and sgnðxÞ_ is the sign function.
Because sgnðxÞ_ is a nonlinear function of x,
_ Eq. (1.1.2) is a nonlinear model.

Time-Invariant Models versus Time-Variant Models

A time-invariant model is one with its governing differential equations having constant
coefficients. A time-variant model is one with its governing differential equations containing
coefficients that change with time. For example, Eq. (1.2.1), with constant coefficients m, c,
and k, describes a time-invariant model. As another example, the differential equation

x þ cx_ þ ðk0 þ ε sin Ω tÞx ¼ f

m€ (1.2.3)

with the time-varying coefficient k0 þ ε sin Ω t, gives a time-variant model.

Continuous-Time Models versus Discrete-Time Models

A continuous-time model is one with its variables as continuous functions time, and it is
described by differential equations. A discrete-time model is one with its variables only
specified at discrete time points, and it is governed by difference equations. Discrete-time
models are usually considered in automatic control and signal processing with digital
computers, in which quantization of signals is performed.
It should be noted that for the same physical system, different types of mathematical
models can be adopted, depending on the application and interest. For instance, in studying
the dynamic behaviors of an airplane, a lumped model may be good enough to describe the
flight trajectory and motion of the airplane in space. However, a distributed model is
necessary to understand the vibration and stress in the airplane structure, for performance
evaluation and failure analysis.
Once a specific model type is selected in design or investigation, the words “model” and
“system” are often used without distinction. Thus, for convenience of discussion, terms like
lumped system, time-invariant system, time-variant system, and nonlinear system mean the
relevant mathematical models.
In this book, lumped and continuous-time systems are considered in modeling, simulation,
and analysis. The coverage of distributed systems and discrete-time systems, which requires
advanced mathematics in modeling and solution, is beyond the scope of this undergraduate-
level text.
6 1 Introduction

1.3 System Model Representations

In simulation and analysis of general dynamic systems, a mathematical model that is
obtained from the first laws of nature is usually converted to an equivalent model represen-
tation. In this book, four commonly used forms of system model representation are intro-
duced: (i) transfer function formulation, (ii) state-space representation, (iii) block diagram
representation in the s-domain, and (iv) block diagram representation in the time domain.
The purpose of adopting these forms is three-fold: to have a standard form for numerical
solutions, to understand the interactions among the components of a system, and to obtain
the information about the internal variables (state variables) of a system. As a matter of fact,
software packages MATLAB/Simulink and Mathematica are created based on these model-
representation forms.
In the subsequent subsections, the above-mentioned system model representations are
illustrated on simple models of mechanical systems, although the basic concepts are applicable
to general dynamic systems.

1.3.1 Transfer function Formulation

Consider a spring–mass–damper system shown in Figure 1.3.1, which is a simple mathemat-
ical model for many mechanical systems. By Newton’s laws and through use of a free-body
diagram of the mass (Section 3.2), the equation of the motion governing the displacement
x of the system is obtained and it is given in Eq. (1.2.1).
Let the input and output of the system be the external force f and the displacement x of the
system, respectively. The transfer function GðsÞ of a dynamic system is defined as the
following input-to-output ratio


L½output


GðsÞ ¼
(1.3.1)
L½input

zero initial conditions

where L is the Laplace transform operator, s is the complex Laplace transform parameter,
and zero initial conditions mean all the initial values of the output and their time derivatives

x (t)
k

m f (t)

Figure 1.3.1 A spring–mass–damper system

 1.3 System Model Representations 7

are set to zero. (Refer to Section 2.4 for the Laplace transform.) Perform a Laplace transform
of Eq. (1.2.1) with respect to time and use zero initial conditions, to obtain
 
ms2 þ cs þ k X ðsÞ ¼ FðsÞ (1.3.2)

where X ðsÞ and FðsÞ are the Laplace transforms of the output x and the input f, respectively.
According to the definition (1.3.1), the transfer function of the spring–mass–damper system
is

X ðsÞ 1
GðsÞ ¼ ¼ 2 (1.3.3)
FðsÞ ms þ cs þ k

Equation (1.3.3) gives an input–output relation (transfer function formulation):

1
X ðsÞ ¼ GðsÞFðsÞ ¼ FðsÞ (1.3.4)
ms2 þ cs þ k

Thus, in the s-domain (the Laplace transform domain), the system output is the product of
its transfer function and the input.
The transfer function formulation (1.3.4) can be used to study the dynamic behavior of the
mechanical system, to determine the system response, and to perform system response
analysis. With the same concept, a transfer function formulation can be established to
a dynamic system of multiple components, by which the dynamic interactions among the
components can be investigated and the system-level analysis can be carried out.
The transfer function formulation for mechanical, electrical, thermal, and fluid systems is
introduced in Chapters 3 to 5, and its application in multi-component dynamic systems and
feedback control systems is covered in Chapters 6 to 8.
It should be pointed out that transfer function formulation is only valid for linear time-
invariant systems and it is not applicable to time-variant systems and nonlinear systems. This
is because the Laplace transform operator is a linear operator with its utility limited to linear
equations with constant coefficients.

1.3.2 State-Space Representation

State-space representation is a modeling technique that is applicable to general dynamic
systems, including time-variant systems and nonlinear systems.
A state-space representation consists of two sets of equations in the time domain: state
equations and output equations. State equations are a set of coupled first-order differential
equations about quantities known as state variables. Output equations are a set of coupled
algebraic equations about the state variables. Here, state variables are used to describe the
mathematical status of a dynamic system, which is known as the state, and by which the
future behavior of the system can be determined with known inputs.
8 1 Introduction

As an example, consider the same spring–mass–damper system as shown in Figure 1.3.1. By

selecting two state variables

x1 ¼ x; x2 ¼ x_ (1.3.5)

the differential equation (1.2.1) can be converted to the following two state equations

x_ 1 ¼ x2
1h i (1.3.6)
x_ 2 ¼  kx1  cx2 þ f
m

Here, the number of state variables is the same as the order of the original differential equation,
which is two for the spring–mass–damper system. Refer to Section 3.7 on how to select state
variables. The state variables are also known as internal variables, which can completely
describe the dynamic behavior of the system, including system outputs and internal forces.
The output equations depend on the selection of system outputs. For instance, if the
velocity and spring force are chosen as the outputs, the output equations of the system are of
the algebraic form

y1 ¼ x2
(1.3.7)
y2 ¼ kx1

where y1 and y2 are the system outputs. As can be seen from Eq. (1.3.7), the outputs are
expressed in terms of the state variables.
To show the versatile utility of state-space representation, consider the spring–mass–
damper system subject to dry friction, which is governed by the nonlinear differential
equation (1.2.2). With the same state variables as given in Eq. (1.3.5), the state equations
are obtained as follows

x_ 1 ¼ x2
1h i (1.3.8)
x_ 2 ¼  kx1  cx2  μN sgnðx2 Þ þ f
m

Equations (1.3.8) are nonlinear state equations. Also, for the time-variant system as
described by Eq. (1.2.3), a set of state equations with the state variables defined in Eq.
(1.3.5) are as follows

x_ 1 ¼ x2
1h i (1.3.9)
x_ 2 ¼ ðk0 þ ε sin Ω tÞx1  cx2 þ f
m

For more details on state-space representation, refer to Sections 2.5 and 6.2, and the
examples in Chapters 3 to 5.
 1.3 System Model Representations 9

State-space representation provides a platform for numerical simulation of the response of

dynamic systems. For instance, the state equations (1.3.6), (1.3.8), and (1.3.9) can be solved
by standard numerical integration algorithms (see Section 2.8). As a matter of fact, some
differential-equation solvers of MATLAB and Mathematica are developed based on state-
space representation.

1.3.3 Block Diagram Representation in the s-Domain

Block diagrams are widely used in representing dynamic systems in control engineering.
A block diagram is an assembly of blocks, each of which describes an input–output relation
for a component or subsystem. In the s-domain, every block is identified with a transfer
function of the component. For instance, for the spring–mass–damper system in
Figure 1.3.1, its block is shown in Figure 1.3.2, where the block is a box having input FðsÞ
and output X ðsÞ, and inside the box is the transfer function of the system. The input–output
relation of the block diagram in general is

Output ¼ Transfer function  Input (1.3.10)

By this relation, the block diagram in Figure 1.3.2 indicates

1
X ðsÞ ¼ FðsÞ (1.3.11)
ms2 þ cs þ k

which is equivalent to the transfer function given in Eq. (1.3.3).

To see the interconnection of system components in a diagram, consider a spring–mass–
damper system with two masses, shown in Figure 1.3.3, where x1 and x2 are the displacements
of masses m1 and m2 , respectively. Following a modeling technique described in Section 3.2,
the governing equations of motion of the system are obtained as follows

m1 x€1 þ cx_ 1 þ ðk þ k1 Þx1  k1 x2 ¼ 0

(1.3.12)
m2 x€2 þ k1 x2  k1 x1 ¼ f

Taking the Laplace transform of Eq. (1.3.12) with zero initial conditions gives the s-domain
equations

ðm1 s2 þ cs þ kÞX1 ðsÞ ¼ k1 ðX2 ðsÞ  X1 ðsÞÞ

(1.3.13)
m2 s2 X2 ðsÞ ¼ FðsÞ  k1 ðX2 ðsÞ  X1 ðsÞÞ

F(s) 1 X (s)

m2 + cs + k

Figure 1.3.2 Block diagram of a spring–mass–damper system

10 1 Introduction

x1 x2
k
k1
m1 m2 f

Figure 1.3.3 A mechanical system with two masses

F (s) – 1 X2 (s) 1 X1(s)

+ +
k1
m2 s2 m1 s2 + cs + k
– Fs (s)

Figure 1.3.4 An s-domain block diagram of the two-mass mechanical system in Figure 1.3.3

Let the external force f and the displacement x1 be the input and output of the system,
respectively. By applying the rules given in Section 6.2.3 to Eq. (1.3.13), the block diagram of
the two-mass system is constructed in Figure 1.3.4, where the right box (for m1 ) and the left
box (for m2 ) are connected through use of a summation point and a pick-off point; and Fs ðsÞ
 
is the internal force of the spring that couples the two masses by Fs ðsÞ ¼ k1 X2 ðsÞ  X1 ðsÞ .
The construction of the block diagram in Figure 1.3.4 is detailed in Section 3.6.
Note that the block diagram representation in the s-domain is limited to linear time-
invariant systems. This is because the subsystem blocks in such a diagram are created
through use of transfer functions.

1.3.4 Block Diagram Representation in the Time Domain

Like state-space representation, block diagram representation in the time domain is applic-
able to general dynamic systems with time-dependent coefficients and nonlinearities. Time-
domain block diagram representation differs from state-space representation in that it
provides a graphical way to model multi-component systems. Learning time-domain block
diagrams helps one to use the software package MATLAB/Simulink in the modeling and
simulation of complicated systems. Indeed, the structure of time-domain block diagrams and
that of Simulink models are quite similar (see Sections 6.5 and 6.6).
 1.4 Analogous Relations 11

f + +
∫ ∫
x x x
1/m
– –

+
c
+

μN sgn

Figure 1.3.5 A time-domain block diagram of the nonlinear system described by Eq. (1.2.2)

As an example, consider the nonlinear differential equation (1.2.2), which models

a spring–mass–damper system subject to dry friction. Let f and x be the input and output
of the system, respectively. By following the steps in Section 6.5, a time-domain block
diagram is constructed in Figure 1.3.5. Refer to Table 6.5.1 for the definitions of the symbols
used in the diagram. Based on the time-domain block diagram, a Simulink model for the
nonlinear system can be conveniently built and the system response to a given input can be
easily computed by running the model.

1.4 Analogous Relations

In this text, four types of dynamic systems are modeled: mechanical systems (Chapter 3),
electrical systems (Chapter 4), and thermal and fluid systems (Chapter 5). As mentioned in
Section 1.1.2, one of the three important keys in modeling is the usage of basic elements,
which are approximate models of components or subsystems and which are the building
blocks of a dynamic system. For this, three types of basic lumped-parameter elements are
introduced: capacitance, resistance, and inertance. For instance, springs, dampers, and masses
in mechanical systems are capacitance, resistance, and inertance elements; capacitors, resis-
tors, and inductors in electric circuits are also capacitance, resistance, and inertance elements.

Flow and Effort

Capacitance, resistance, and inertance elements for mechanical, electrical, thermal, and fluid
systems can be similarly treated by equations involving two variables: flow q and effort f.
Here, q can be a velocity, a current, heat flow rate, or a mass flow rate while f can be an
12 1 Introduction

external force, a voltage, a temperature, or a pressure. Physically, the effort of an element is

related to the application of an action to the element, which is an input to the element, and
the flow is about the consequence of the action, which is an output of the element.
Flow and effort are also known as a pair of power variables because the power P of an
element can be expressed by

P¼f q (1.4.1)

For instance, for an electrical element (resistor or capacitor or inductor), its electrical power
is
P¼ vi

where v is the voltage (potential drop) of the element, and i is the current going through it. In
this case, current i and voltage v are the effort and flow of the electrical element, respectively.
The power f  q results in storage of potential energy or kinetic energy in the element or in
dissipation of energy by the element. As another example, the power of a mechanical element
is

P¼f v

where f is the force applied to the element, and v ¼ x_ is the velocity of the element. In this
case, velocity v and force f are the effort and flow of the mechanical element, respectively.

Element Laws
e R
Denote capacitance, resistance, and inertance of general elements by C, e and eI , respectively.
The governing equations for general elements are stated as follows:

e df
For capacitance : q ¼ C (1.4.2)
dt
e
For resistance : f ¼ Rq (1.4.3)
dq
For inertance : f ¼ eI (1.4.4)
dt

where q and f are flow and effort, respectively.

Equations (1.4.2) to (1.4.4) are known as the element laws that characterize analogous
e R,
relations among different types of systems by the equivalent coefficients C, e and eI . In
general, a capacitance element stores potential energy; a resistance element dissipates energy;
and an inertance element stores kinetic energy.
Consider the spring–mass–damper system in Figure 1.3.1, where the spring (k), mass (m),
and damper (c) are capacitance, resistance, and inertance elements. In what follows, we show
 1.5 Software for Computation 13

that the physical parameters of the system are related to the capacitance, resistance, and
inertance by

1
Ce ¼ ; e ¼ c;
R eI ¼ m (1.4.5)
k
e and eI in Eq. (1.4.5) can be easily obtained from the following equations
The R

fd ¼ cx_ ¼ cv (1.4.6)
x ¼ mv_ ¼ f
m€ (1.4.7)

where fd is the internal force (damping force) of the damper, and Eq. (1.4.7) is the momentum
equation for the mass according to Newton’s second law. The internal force (spring force) fs
of the spring can be written as fs ¼ kx, which after differentiation leads to

dfs
¼ kx_ ¼ kv (1.4.8)
dt

It follows that

1 dfd
v¼ (1.4.9)
k dt
e ¼ 1=k:
indicating that C
In summary of the previous discussion, the analogous variables and equivalent coefficients
of lumped parameter elements of mechanical, electrical, thermal, and fluid systems are listed
in Table 1.4.1. In the table, a spring–mass–damper system is used as a representative of
mechanical systems and a hydraulic system is used as a representative of fluid systems. Note
that thermal systems, by the law of thermodynamics, do not have inertance elements. These
analogous relations are helpful for better understanding of the element laws, and they are
useful in model development for general dynamic systems.

1.5 Software for Computation

All the numerical results presented in this text are obtained through use of MATLAB,
Simulink, and Mathematica. Appendices B and C provide tutorials on these software
packages. In Sections 6.6 and 6.7, use of MATLAB, Simulink, and Mathematica is illus-
trated in examples. Also, several application problems in Chapter 9 are solved by MATLAB,
Simulink, and Mathematica.
Note that numerical solution via the above-mentioned software packages adopts transfer
function formulation, state-space representation, and block diagrams. Therefore, mastering
Table 1.4.1 Analogous variable and equivalent coefficients

Mechanical system Electrical system Thermal system Fluid system

dx dqe
Flow q¼ i¼
dt dt Heat flow rate qh Mass flow rate qm
x ¼ displacement qe ¼ electric charge
Effort f = force v = voltage Temperature T Pressure p
Capacitance Spring: Capacitance: Thermal capacitance: Fluid capacitance:
e ¼1 e ¼C
C e ¼ C ¼ dE e ¼ C ¼ dm
C C C
k dt dp
Resistance Damper: Resistance: Thermal resistance: Fluid resistance:
e ¼c
R e¼R
R e ¼R
R e ¼ R ¼ dp
R
dq
Inertance Mass: Inductance: Fluid inertance:
eI ¼ m eI ¼ L eI ¼ I
 1.6 Units 15

these forms of system model representation is essentially important to modeling, simulation,

and analysis of dynamic systems.

1.6 Units
In this text, we use the International System of Units (SI), which is also known as the metric
system. Refer to Appendix A for the units of commonly used quantities, and a table of unit
conversion between the SI system and the US customary system.
2 Fundamentals of Mathematics

Contents
2.1 Vector Algebra 16
2.2 Matrix Algebra 24
2.3 Complex Numbers 34
2.4 The Laplace Transform 37
2.5 Differential Equations 63
2.6 Solution of Linear Ordinary Differential Equations with Constant
Coefficients 70
2.7 Transfer Functions and Block Diagrams of Time-Invariant Systems 95
2.8 Solution of State Equations via Numerical Integration 98
Chapter Summary 102
References 103
Problems 103

This chapter provides the reader with a brief refresher course on the mathematical apparatus
crucial for modeling of dynamic systems. Sections 2.1 and 2.2 present basic concepts and
terminology of vector and matrix algebra. Definitions and basic operations on complex
numbers are introduced in Section 2.3. Section 2.4 is devoted to one of the important
methods for solving differential equations – the Laplace transform. Sections 2.5 and 2.6
discuss the types of differential equations widely encountered in modeling of common
dynamic systems and develop methods for solving these equations. Section 2.7 introduces
the mathematical foundation for deriving transfer functions and creating block diagrams of
various linear time-invariant dynamic systems. Section 2.8 presents a brief overview of
solving differential equations numerically, with MATLAB and Wolfram Mathematica.

2.1 Vector Algebra

Components of a dynamic system can be represented by different physical quantities. Some
of these quantities such as, for example, mass or moment of inertia, only have magnitude.

16
 2.1 Vector Algebra 17

These quantities are called scalars and are represented by numbers alone with the appropri-
ate units. The other physical quantities such as force, position, and velocity, for example, are
characterized by magnitude and direction; thus, they can be manipulated in a way that takes
into account both mentioned parameters. These quantities are represented by vectors.

2.1.1 Definition
A vector is a mathematical quantity that has both magnitude and direction. It is customarily
depicted as an arrow pointing in the direction of the vector. The length of the arrow
corresponds to the magnitude of the vector.
The origin of the vector is called the tail, and its end is called the head (Figure 2.1.1). Since a
vector is unchanged under a translation – a transformation consisting of a constant offset
with no rotation or distortion – it is often
convenient to consider the vector tail A as
z
located at the origin of the coordinate system.
A vector can be denoted by its tail and head
Head
~ or by a symbol such as ~
such as AB, v: ~
v ¼ AB.~ B
The arrow above the vector symbol may be v
omitted, having a vector denoted as just v. Tail
As a formal mathematical object, a vector is A
y
commonly defined as an element of a vector
space and is specified by its coordinates. In the
Cartesian coordinate system, a vector is x

defined by its three coordinates – x, y, and z

2 3 2 3
vx xB  x A Figure 2.1.1 A vector
– as follows: ~ 4 5 4
v ¼ vy ¼ yB  yA . The 5
vz zB  zA
magnitude of a vector is called the length, or sometimes the norm, and is defined as
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
j~
vj ¼ vx 2 þ v y 2 þ v z 2 (2.1.1)

A vector with unit length is called a unit vector, and is denoted with a hat above its symbol
such as ^v . An arbitrary vector can be converted to a unit vector – normalized – by dividing by
its length: ^v ¼ ~
j~
v . A zero vector, denoted as 0, is a vector of arbitrary direction and length
vj
zero, having all its coordinate components equal to zero.
Vectors can exist at any point in space. Any two vectors that have the same length
and the same direction are considered equal, no matter where in coordinate space they
are located.
18 2 Fundamentals of Mathematics

2.1.2 Algebraic Manipulations: Addition, Subtraction, and Scalar Multiplication

The most important operations on vectors include addition, subtraction, and scalar multi-
plication. Vector multiplication is not uniquely defined and for any arbitrary pair of vectors
includes the dot product and cross product.
Vector addition is performed as shown in Figure 2.1.2(a) – vectors ~
p and~q are placed head
to tail, and the resulting vector sum ~
v ¼~ p þ~q is drawn from the free tail to the free head.
Vector addition is commutative:

~
p þ~
q ¼~
q þ~
p (2.1.2)

and associative:
p þ~
ð~ r ¼~
qÞ þ~ p þ ð~
q þ~
rÞ (2.1.3)

For any arbitrary vector ~

p, a zero vector performs as an identity element for vector addition:

p þ~
~ 0 ¼~
0 þ~
p ¼~
p (2.1.4)

Also, for any vector ~

p, there is a unique inverse vector ð1Þ~ p ≡ ~ p such that
~
p þ ð~ ~
pÞ ¼ 0. That means that vectors ~
p and ~ pj ¼ j  ~
p have the same length, j~ pj, but
opposite orientations.
Vector subtraction is illustrated in Figure 2.1.2(b). It is an inverse operation to vector addition,
and is performed as addition with the orientation of the second vector reversed, such as

~
v ¼~
p ~
q ¼~
p þ ð~
qÞ (2.1.5)

When vectors are defined by their coordinates, vector sums resulting from addition/
2 3 2 3
px þ qx px  qx
vaddition ¼ 4 py þ qy 5 and ~
subtraction are: ~ v subtraction ¼ 4 py  qy 5
pz þ qz pz  qz

Figure 2.1.2 (a) Addition and (b) subtraction of vectors

 2.1 Vector Algebra 19

Scalar multiplication is the operation of multiplying a vector by some constant a. If

~
v ¼ a ~p, then vectors ~v and ~
p are of the same direction if a > 0 and of opposite direction
2 3
if a < 0, while length of the vector ~
v is j~
vj ¼ a  j~
pj. a  px
In coordinate notation, the vector ~ v ¼ 4a  py 5.
v is represented as ~
Scalar multiplication is commutative: a  pz

a ~
p ¼~
pa (2.1.6)

distributive:
ða þ bÞ  ~
p ¼ a ~
p þ b ~
p
p þ~
a  ð~ qÞ ¼ a  ~
p þ a ~
q (2.1.7)

and associative:
b  ða  ~
pÞ ¼ ðb  aÞ  ~
p ¼ ða  bÞ  ~
p ¼ a  ðb  ~
pÞ (2.1.8)

The number 1 acts an identity element for scalar multiplication:

1 ~
p ¼~
p

2.1.3 Unit-Vector Decomposition

For a chosen coordinate system with an origin and
axes, any vector can be decomposed into compo-
nent vectors along each coordinate axis, as shown
in Figure 2.1.3.
An arbitrary vector ~ v in the Cartesian three-
dimensional space can be decomposed into the fol-
lowing vector sum: ~ v ¼~ v x þ~ v z where ~
v y þ~ vx, ~
vy,
and ~v z are vector components parallel to the x-, y-,
and z- coordinate axes, respectively. Let us define
unit vectors ^i, ^j, and ^k collinear with the x-, y-, and
Figure 2.1.3 Unit-vector
decomposition
z- coordinate axes, respectively. As mentioned earl-
ier, a unit vector means that its magnitude is one, that is j^ij ¼ j ^jj ¼ j^kj ¼ 1. As indicated in
Figure 2.1.3, the direction of the unit vector ^i points in the direction of the increasing x-
coordinate, i.e. it is pointing in the +x direction. Directions of unit vectors ^j and ^k are
similarly defined. Having defined the unit vectors, the vector components and the decom-
position itself can be expressed as follows:

v x ¼ vx~i
~

v y ¼ vy~j
~
20 2 Fundamentals of Mathematics

v z ¼ vz~
~ k

v ¼ vx~i þ vy~j þ vz~

~ k (2.1.9)

In the expressions above, the terms vx , vy , and vz are called the x-, y-, and z- components of
vx, ~
the vectors ~ v y , and ~
v z respectively. While the absolute value of the term vx is equal to the
magnitude of the vector ~ v x , vx itself can be either positive or negative. Recalling the vector
definition in coordinate notation it can be stated that vx is the x- coordinate of vector ~ v. The
same reasoning applies to the terms vy and vz .
In terms of unit vector decomposition, vector addition, subtraction, and scalar multipli-
cation can be expressed as

~ q ¼ ðpx ^i þ py ^j þ pz ^kÞ þ ðqx ^i þ qy ^j þ qz ^kÞ

p þ~

¼ ðpx þ qx Þ ^i þ ðpy þ qy Þ ^j þ ðpz þ qz Þ ^k (2.1.10)

~ q ¼ ðpx ^i þ py ^j þ pz ^kÞ  ðqx ^i þ qy ^j þ qz ^kÞ

p ~

¼ ðpx  qx Þ ^i þ ðpy  qy Þ ^j þ ðpz  qz Þ ^k (2.1.11)

p ¼ a  ðpx ^i þ py ^j þ pz ^kÞ ¼ a  px ^i þ a  py ^j þ a  pz ^k
a ~ (2.1.12)

Example 2.1.1 Algebraic operations on vectors, defined by their coordinates

Given two vectors ~
p and ~
q, defined by their coordinates, compute their sum and difference, using
unit-vector decomposition. Also, compute multiplication of vector ~
p by the scalar a ¼ 2.
Solution For two vectors
2 3 2 3
1:0 3:2
p ¼ 4 2:1 5 and ~
~ q ¼ 41:1 5
4:7 3:2

the sum is computed by Eq. (2.1.10) as

~ q ¼ ð1:0 ^i þ 2:1 ^j þ 4:7 ^kÞ þ ð3:2 ^i  1:1 ^j  3:2 ^kÞ

p þ~

¼ ð1:0 þ 3:2Þ ^i þ ð2:1  1:1Þ ^j þ ð4:7  3:2Þ ^k

¼ 2:2 ^i þ 1:0 ^j þ 1:5 ^k

while the difference is computed by Eq. (2.1.11) as

 2.1 Vector Algebra 21

~ q ¼ ð1:0 ^i þ 2:1 ^j þ 4:7 ^kÞ  ð3:2 ^i  1:1 ^j  3:2 ^kÞ

p ~

¼ ð1:0  3:2Þ ^i þ ð2:1  ð1:1ÞÞ ^j þ ð4:7  ð3:2ÞÞ ^k

¼ 4:2 ^i þ 3:2 ^j þ 7:9 ^k

Multiplication of vector ~
p by the scalar a is performed by Eq. (2.1.12) as

p ¼ 2:0  ð1:0 ^i þ 2:1 ^j þ 4:7 ^kÞ

a ~

¼ 2:0  ð1:0Þ ^i þ 2:0  2:1 ^j þ 2:0  4:7 ^k

¼ 2:0 ^i þ 4:2 ^j þ 9:4 ^k

2.1.4 Vector Multiplication Operations

Vector multiplication operations such as the dot
product and cross product, as well as the right-
hand rule, are illustrated in Figure 2.1.4.
The dot-product operation is often called the
v = p ×q
scalar product. It takes two vectors and generates
a scalar quantity – a single number. The dot
product ~p ~
q of two vectors ~ p and ~
q is defined as p
a product of their magnitudes with the cosine of θ
the angle θ between them (0 ≤ θ ≤ π): q

~
p ~
q ¼ j~
pj j~
qjcos θ (2.1.13)
Figure 2.1.4 The right-hand rule
Obviously, depending on the value of the cos θ
the dot product can be positive (a sharp angle
between two vectors), negative (an obtuse angle between two vectors), and zero (the vectors
are perpendicular).
The dot product is commutative:
~
p ~
q ¼~
q ~
p (2.1.14)

and distributive:
ð~ rÞ  ~
p þ~ q ¼~
p ~ r ~
q þ~ q
~
q  ð~ rÞ ¼ ~
p þ~ q ~
p þ~
q ~
r (2.1.15)

With regard to scalar multiplication the following properties hold:

ða  ~
p Þ ~ p ~
q ¼ a  ð~ qÞ
~
p  ða  ~ p ~
qÞ ¼ a  ð~ qÞ (2.1.16)
22 2 Fundamentals of Mathematics

If vectors of the dot-product operation are defined by their unit-vector decompositions, i.e.
they are defined in terms of coordinates, the dot product becomes:

~ q ¼ ðpx^i þ py^j þ pz ^kÞ  ðqx^i þ qy^j þ qz ^kÞ

p ~ (2.1.17)

To simplify this expression, the dot products of unit vectors need to be derived. Using the
dot-product definition, it can be shown that the dot product of the unit vector with itself is
unity: ^i  ^i ¼ j^ij j^ijcosð0Þ ¼ 1. Similarly, ^j  ^j ¼ 1 and ^k  ^k ¼ 1. Since two distinct unit vectors
(as defined collinear to the coordinate axes) are perpendicular to each other, their dot
π 
product is zero: ^i  ^j ¼ j^i j j ^j jcos ¼ 0. Similarly, ^i  ^k ¼ 0 and ^j  ^k ¼ 0. Thus, Eq. (2.1.17)
2
yields:

~ q ¼ ðpx^i þ py^j þ pz ^kÞ  ðqx^i þ qy^j þ qz ^kÞ

p ~
¼ px qx þ py qy þ pz qz (2.1.18)

The cross product also takes a pair of vectors, but instead of a scalar it generates a new
vector. If some vector ~ v represents the cross product of two vectors ~ p and ~q with the angle θ
between them (0 ≤ θ ≤ π), i.e. ~
v ¼~ p ~q, then the direction of this vector ~
v is perpendicular to
the plane formed by ~ p and ~ q (as shown in Figure 2.1.4), and its magnitude is
j~ p ~
vj ¼ j~ qj ¼ j~
pj j~
qjsin θ.
The direction of vector ~ v ¼~ p ~ q is determined using the so-called right-hand rule. It
reads as follows: if the curl of the right-hand fingers represents a rotation from ~ p to ~
q, then~
v
points in the direction of the outstretched right thumb. This is illustrated in Figure 2.1.4.
The cross product is anti-commutative, since changing the order of the vectors in the cross
product reverses the direction of the resulting vector by the right-hand rule:

~
p ~ q ~
q ¼ ~ p (2.1.19)

The cross product is distributive:

ð~ rÞ  ~
p þ~ q ¼~
p ~ r ~
q þ~ q
~
q  ð~ rÞ ¼ ~
p þ~ q ~
p þ~
q ~
r (2.1.20)

With regard to scalar multiplication the following properties hold:

ða  ~
p Þ ~ p ~
q ¼ a  ð~ qÞ
~
p  ða  ~ p ~
qÞ ¼ a  ð~ qÞ (2.1.21)

To derive an expression for the cross product of vectors defined by their unit-vector
decompositions, cross products of unit-vector pairs need to be evaluated. Using the defin-
ition of the cross product, we obtain the following.
 2.1 Vector Algebra 23

• The cross product of a unit vector with itself is zero:

^i  ^i ¼ j^ij j^ij sinð0Þ ¼ 0 (2.1.22)

Similarly, ^j  ^j ¼ 0 and ^k  ^k ¼ 0.
• The cross product of two distinct unit vectors is computed as follows:

^i  ^j ¼ ^k (2.1.23)

• The magnitude of the cross product ^i  ^j is unity:

 
^i  ^j ¼ j^i j j^j j sin π ¼ 1 (2.1.24)
2

According to the right-hand rule, the direction of the cross product ^i  ^j is þ^k. Thus, Eq.
(2.1.24) holds true.
Similarly, ^j  ^k ¼ ^i and ^k  ^i ¼ ^j. By the anti-commutative property of the cross product
^j ^i ¼ ^k and ^i  ^k ¼ ^j.
Considering the above conclusions, we obtain

~ q ¼ ðpx^i þ py^j þ pz ^kÞ  ðqx^i þ qy^j þ qz ^kÞ

p ~
¼ ðpy qz  pz qy Þ ^i þ ðpz qx  px qz Þ ^j þ ðpx qy  py qx Þ ^k (2.1.25)

Example 2.1.2 Computing the dot product and cross product of two vectors
Given two vectors ~p and ~q, defined by their coordinates, compute their dot product and cross
product, using unit-vector decomposition.
2 3 2 3
Solution 1:0 3:2
p ¼ 4 2:1 5and ~
For two vectors ~ q ¼ 41:1 5 the dot product is computed by Eq. (2.1.18) as
4:7 3:2
~ q ¼ ð1:0 ^i þ 2:1 ^j þ 4:7 ^kÞ  ð3:2 ^i  1:1 ^j  3:2 ^kÞ
p ~
¼ 1:0  3:2 þ 2:1  ð1:1Þ þ 4:7  ð3:2Þ
¼ 20:55

Their cross product is computed by Eq. (2.1.25) as

~ q ¼ ð1:0 ^i þ 2:1 ^j þ 4:7 ^kÞ  ð3:2 ^i  1:1 ^j  3:2 ^kÞ

p ~
¼ ð2:1  ð3:2Þ  4:7  ð1:1ÞÞ ^i þ ð4:7  3:2  ð1:0Þ  ð3:2ÞÞ ^j

þ ðð1:0Þ  ð1:1Þ  2:1  3:2Þ ^k

¼ 1:55 ^i þ 11:84 ^j  5:62 ^k

24 2 Fundamentals of Mathematics

2.2 Matrix Algebra

Matrices are an important element of linear algebra, and are extremely useful in dealing with
the systems of equations that are heavily coupled.

2.2.1 Definition
A matrix is a rectangular array of numbers,
a11 a1n
symbols, or expressions arranged in rows and
columns. Matrices are often shown enclosed in A= m rows
am1 amn
square brackets and their elements are denoted
by their indices that represent an element
n columns
address in the matrix. This is illustrated in
Figure 2.2.1. Figure 2.2.1 A matrix
Matrix A, depicted in Figure 2.2.1, has m rows
and n columns. It is said to be of size m  n.
Obviously, the indices of the last element of the
matrix identify the matrix size. When m ¼ n, the matrix is square. Otherwise, it is rectangu-
lar. An element aij of matrix A is called the (i, j) entry and it is located at the intersection of
the i-th row and the j-th column. In the matrix notation, (i, j) is known as the address of the
element aij in matrix A.
A vector is sometimes called an instance of a matrix; for example, a matrix of a size m  1 is
a column vector, while a matrix of a size 1  n is a row vector.
The abbreviated matrix notation represents matrix A of size m  n as A ¼ ½aij mn or
simply as Amn .
The element the indices of which are the same, i.e. aii , is called a diagonal element. A square
matrix, for which the following is true: aij ¼ Ci 8 i ¼ j and aij ¼ 0 8 i ≠ j, where Ci is a number,
symbol, or expression, is called a diagonal matrix, and is sometimes denoted as
A ¼ Diag½aij mn .
A square diagonal matrix for which aij ¼ 1 8 i ¼ j and aij ¼ 0 8 i ≠ j, is called an identity
matrix, and is denoted Inn . In some literature identity matrix may be also called unit matrix.
A zero matrix or null matrix is an m  n matrix of all zeros.

2.2.2 Single-Matrix Operations

Matrices can undergo several important operations. A single matrix can be transposed,
inverted, and multiplied by a scalar. It can also experience the elementary row and column
operations such as:
 2.2 Matrix Algebra 25

(a) interchanging two rows or columns

(b) multiplying any row or column by a nonzero constant
(c) adding a row or column to another
(d) multiplying the i-th row by a nonzero constant, adding the result to the j-th row, and
replacing the j-th row with the result of addition

Scalar multiplication with regard to a matrix is the operation of multiplying that matrix by
some constant γ. It is done by multiplying each element of the original matrix by the said
constant γ:
0 1 0 1
a11 . . . a1n γ  a11 . . . γ  a1n
B . .. C B .. .. C
γ  @ .. ⋱ . A¼@ . γa . A
ij
(2.2.1)
am1 ... amn γ  am1 ... γ  amn

The matrix transpose is denoted AT . It replaces all elements aij with aji . Hence, rows
become columns, columns become rows, changing the size of the matrix from m  n to
n  m: ½Amn T ¼ Anm . The matrix is called symmetric if AT ¼ A.
The matrix inverse exists only for square matrices. The inverse of matrix Ann is a matrix
½Ann 1 such that A  A1 ¼ I, i.e. the product of a square matrix and its inverse is an identity
matrix.

Example 2.2.1 Multiplication of a matrix by a scalar, and its transpose


1:0 2:1 4:7
Given the matrix A ¼ , multiply it by γ ¼ 1:2 and find the transpose of the
3:2 1:1 3:2
resulting matrix.
Solution
Multiplication of the given matrix by a scalar is done by Eq. (2.2.1) as
 
1:2  ð1:0Þ 1:2  2:1 1:2  4:7 1:2 2:52 5:64
γA¼ ¼
1:2  3:2 1:2  ð1:1Þ 1:2  ð3:2Þ 3:84 1:32 3:84
0 1
1:2 3:84
Transposing the resulting matrix yields ðγ  AÞT ¼ @ 2:52 1:32 A
5:64 3:84

2.2.3 Matrix Determinants

Computation of the matrix inverse involves a concept of matrix determinants. A determin-
ant is a mathematical object, defined only for square matrices. It is denoted as Detð AÞ or jAj.
For a 2  2 matrix, the determinant is easily computable as
26 2 Fundamentals of Mathematics







a a12
a a12

Det 11 ≡
11
≡ a11 a22  a12 a21 (2.2.2)
a21 a22
a21 a22

For larger matrices, computation of the determinant involves expansion by minors. Mij
denotes a minor of the square matrix Ann ¼ ½ai j nn . This minor is a matrix of the size
ðn  1Þ  ðn  1Þ formed by eliminating row i and column j from the original matrix A. The
expansion of the determinant by minors for an arbitrary square matrix Ann is as follows:
0 1
a11 a12 . . . a1n
Ba21 a22 . . . a2n C
B C
DetðAÞ ¼ jAj ¼ DetB .. .. .. C
@ . . ⋱ . A
an1 an2 ... ann








a22 a23 ... a2n


a21 a23 ... a2n


a21 a23 ... a2ðn1Þ







a32 a33 ... a3n


a31 a33 ... a3n


a31 a33 ... a3ðn1Þ






¼ a11
.. .. ..
a12
.. .. ..
þ . . .  a1n
.. .. ..



. . ⋱ .


. . ⋱ .


. . ⋱ .






an2 an3 ... ann

an1 an3 ... ann

an1 an3 ... anðn1Þ

(2.2.3)

Thus, the determinant of an arbitrary matrix Ann is

Xn
jAnn j ¼ i¼1
ai j Ci j (2.2.4)

where Cij ≡ ð1Þiþj Mij . Cij is called the cofactor of aij .

The determinant is additive:







n



 Ann
¼ ð1Þ
Ann
(2.2.5)

and distributive:















Ann  Bnn
¼
Ann

Bnn
(2.2.6)

With regard to scalar multiplication, the determinant is

a  Ann
¼ an
Ann
(2.2.7)

The determinant of the matrix inverse is

1
1

ðAnn Þ
¼




(2.2.8)

Ann

The determinant of a transpose equals the determinant of the original matrix:

T




ðAnn Þ
¼
Ann
(2.2.9)
 2.2 Matrix Algebra 27

The determinant is invariant under elementary row and column operations. The other
useful properties of determinant are:

(a) switching two rows or columns changes the sign of the determinant
(b) scalars can be factored out from rows and columns
(c) scalar multiplication of a row by a constant multiplies the determinant by that constant
(d) determinant of a matrix with a row or column of zeros equals zero
(e) determinant of a matrix with two equal rows or two equal columns is zero

It can easily be shown that the determinant of a triangular or diagonal matrix equals a
product of the elements on its main diagonal. For a 3 × 3 upper triangular matrix, the
determinant is computed as follows:

Example 2.2.2 Finding the matrix determinant

Given matrices

1:0 2:1
A¼
3:2 1:1

and
0 1
1:0 0:6 2:1
B ¼ @ 3:2 4:7 1:1 A
2:5 1:5 1:0

find their determinants.

Solution
The determinant of matrix A can be found by Eq. (2.2.2) as

1:0 2:1
Det ¼ ð1:0Þ  ð1:1Þ  2:1  3:2 ¼ 5:62
3:2 1:1

while to compute the determinant of matrix B we need to use Eq. (2.2.3) that yields
0 1
1:0 0:6 2:1
Det@ 3:2 4:7 1:1 A
2:5 1:5 1:0















4:7 1:1

3:2 1:1

3:2 4:7

¼ 1:0 

0:6 

þ2:1 

¼  41:375

1:5 1:0

2:5 1:0

2:5 1:5

28 2 Fundamentals of Mathematics

0 1






a11 a12 a13







a a23

0 a23

0 a22

Det@ 0 a22 a23 A ¼ a11 
22
a12 

þa13 

¼ a11 a22 a33

0 a33

0 a33

0 0

0 0 a33

For a diagonal matrix, computation of the determinant is even more straightforward:

0 1






a11 0 0







a 0

0 0

0 a22

Det@ 0 a22 0 A ¼ a11 
22
0 

þ0 

¼ a11 a22 a33

0 a33

0 a33

0 0

0 0 a33

The determinant is very useful for solving linear systems of equations. Cramer’s rule
derives a solution of a linear system of equations as quotients of determinants. While
impractical for the larger systems of equations, it is useful for systems of two and three
equations. It also is of interest in differential equations and some engineering applications.
Cramer’s rule states the following. Consider a linear system of n equations in the same
number of unknowns:
2 3
a11 x1 þ a12 x2 þ . . . þ a1n xn ¼ b1
6a21 x1 þ a22 x2 þ . . . þ a2n xn ¼ b2 7
6 7
6 .. 7 (2.2.10)
4 . 5
an1 x1 þ an2 x2 þ . . . þ ann xn ¼ bn

or in matrix form
A :~x ¼~
b
0 1
a11 . . . a1n
B .. .. C
where A is a matrix of coefficients; A ¼ @ . ⋱ . A
an1 . . . ann

~ x ¼ ½x1 x2 . . . xn  T
x is a vector of unknowns; ~

and ~
b is a vector of the results; ~
b ¼ ½b1 b2 . . . bn  T

Then, if the determinant of matrix A is nonzero:

D ¼ DetðAÞ ≠ 0 (2.2.11)

the given linear system of equations has exactly one solution, derived as follows (this
expression is typically referred to as Cramer’s rule):
D1 D2 Dn
x1 ¼ ; x2 ¼ ; . . . xn ¼ (2.2.12)
D D D

where Dk is obtained from D by replacing the k-th column in D with the vector ~ b: If the
~
system is homogeneous, i.e. b ¼ 0, and D ≠ 0, then this system has only the trivial solution
x1 ¼ x2 ¼ . . . ¼ xn ¼ 0. If the homogeneous system has a zero coefficient determinant
(D ¼ 0), then it has also nontrivial solutions.
 2.2 Matrix Algebra 29

Example 2.2.3 Solving a linear system of equations using Cramer’s rule

x ¼~
Consider a linear system defined by the matrix equation A : ~ b, where matrix

1:0 2:1
A¼ x ¼ ½x1 x2  T , and ~
,~ b ¼ ½3:2 2:1 T .
3:2 1:1

Use Cramer’s rule (Eq. (2.2.11)) to find out whether a unique solution exists.
Solution

D ¼ DetðAÞ ¼ 1:0  ð1:1Þ  2:1  3:2 ¼ 5:62 ≠ 0

Since the determinant of coefficients matrix is nonzero, the given system has exactly one
solution, derived with Cramer’s rule:




3:2 2:1




D1
2:1 1:1
7:93
x1 ¼ ¼ ¼ ¼ 1:41
D 5:62 5:62




1:0 3:2




D2
3:2 2:1
12:34
x2 ¼ ¼ ¼ ¼ 2:196
D 5:62 5:62

This solution can easily be verified by writing the given system in the expanded form:

1:0x1 þ 2:1x2 ¼ 3:2

3:2x1  1:1x2 ¼ 2:1

deriving the expression for x1 from the first equation, plugging it into the second equation,
deriving the solution for x2 and subsequently for x1 :

x1 ¼ 2:1x2  3:2
3:2ð2:1x2  3:2Þ  1:1x2 ¼ 2:1 )
2:1 þ 3:2  3:2
x2 ¼ ¼ 2:196 ) x1 ¼ 2:1  2:196  3:2 ¼ 1:41
3:2  2:1  1:1

Proof that a homogeneous system

1:0x1 þ 2:1x2 ¼ 0
3:2x1  1:1x2 ¼ 0

has only a trivial solution is straightforward:

  
1:0 2:1 x 0
This homogeneous system, : 1 ¼ , has multiple nontrivial solutions in
2:0 4:2 x2 0
addition to the trivial one:

D ¼ DetðAÞ ¼ 1:0  4:2  2:1  2:0 ¼ 0

As seen from the coefficient matrix, the second equation of the system is obtained from the first
one by multiplying it by 2. Hence, this linear system can be described by a single equation
30 2 Fundamentals of Mathematics

1:0x1 þ 2:1x2 ¼ 0, which has an infinite number of solutions written as x1 ¼ 2:1x2 with x2 being
1
an independent variable, or as x2 ¼  x1 with x1 being an independent variable.
2:1

2.2.4 Matrix Inverse

The application of a concept of determinants to the computation of a matrix inverse allows
formulating the following condition of the existence of the inverse. An arbitrary square
matrix Ann has an inverse if and only if its determinant is nonzero: jAnn j ≠ 0. If this holds
true, the matrix A is called nonsingular or invertible.
For a 2  2 nonsingular matrix

a11 a12
A¼
a21 a22

an inverse is easily computed:

 
1 a22 a12 1 a22 a12
A1 ¼ ¼ (2.2.13)
jAj a21 a11 a11 a22  a12 a21 a21 a11

For a 3  3 nonsingular matrix

2 3
a11 a12 a13
A ¼ 4a21 a22 a23 5
a31 a32 a33

The computation of an inverse is also straightforward and involves computing determinants

of several 2  2 matrices – minors of matrix A
:

(2.2.14)

In general, an inverse of an arbitrary nonsingular matrix Ann is computed using the

adjoint of matrix A, which is defined as follows:
 2.2 Matrix Algebra 31

2 3
ð1Þ1þ1 M11 ð1Þ2þ1 M21 ... ð1Þnþ1 Mn1
6ð1Þ1þ2 M12 ð1Þ2þ2 M22 ... ð1Þnþ2 Mn2 7
6 7
adjðAÞ ¼ 6 .. .. .. 7 ¼ ½ci j nn (2.2.15)
4 . . ⋱ . 5
ð1Þ1þn M1n ð1Þ2þn M2n ... ð1Þnþn Mnn

where Mij denotes a minor of the square matrix Ann ¼ ½aij nn , and every element
cij ¼ ð1Þiþj Mij of the adjoint matrix is the cofactor of aij . It shall be noted that cij occupies
the ð j; iÞ position in the adjoint matrix.
In the adjoint-matrix formulation the inverse of matrix Ann is
1
A1 ¼ adjðAÞ (2.2.16)
jAj

Inverting an arbitrary nonsingular square matrix Ann involves a fair amount of compu-
tation and is usually done with software such as MATLAB, Mathematica, or Maple.

Example 2.2.4 Computing the matrix inverse

The inverse of matrix

1:0 2:1
A¼
3:2 1:1

is computed by Eq. (2.2.13) as

 
1 1 1:1 2:1 0:196 0:374
A ¼ ¼
jA j 3:2 1:0 0:569 0:178

The inverse of 3 × 3 matrix

0 1
1 1 2
A ¼ @ 3 1 1 A
1 3 4

is computed by Eq. (2.2.14) as

jAj ¼ ð1Þ  ðð1Þ  4  1  3Þ  1  ð3  4  1  ð1ÞÞ þ 2  ð3  3  ð1Þ  ð1ÞÞ ¼ 10
32 2 Fundamentals of Mathematics

2.2.5 Operations on Two Matrices

Two matrices can be added, subtracted, or multiplied together to yield a new matrix. Two
matrices are equal if they have the same size and the same elements in the respective
locations.
The sum of two matrices C ¼ A þ B is defined by adding elements with the same
indices: cij ≡ aij þ bij over all i; j. For addition to be possible, matrices A and B must be
of the same size. Subtraction is a combination of addition and scalar multiplication as
follows:

A  B ¼ A þ ðBÞ (2.2.17)

Matrix addition is commutative:

AþB¼BþA (2.2.18)

and associative:
ðA þ BÞ þ C ¼ A þ ðB þ CÞ (2.2.19)

The transpose of a matrix sum equals the sum of transposes of member matrices:

ð A þ BÞ T ¼ AT þ BT (2.2.20)

Multiplication of two matrices A and B is possible if and only if their dimensions correlate
as ðn  mÞ  ðm  pÞ ¼ ðn  pÞ, where n  m are the dimensions of matrix A, m  p are the
dimensions of matrix B, and n  p are the dimensions of their product C ¼ A  B. The
product of two matrices Cnp ¼ Anm  Bmp is defined as

ci j ¼ ai1 b1j þ ai2 b2j þ ai3 b3j þ . . . þ aim bm j (2.2.21)

An (i, j) element of the product matrix C is the dot (inner) product of the i-th row of
matrix A and the j-th column of matrix B, i.e. a summation of elements of the i-th row
of matrix A multiplied one-by-one by their respective elements of the j-th column of
matrix B.
The matrix multiplication product can be written explicitly as follows:
0 1 0 1 0 1
a11 a12 ... a1m b11 b12 ... b1p c11 c12 ... c1p
Ba21 a22 ... a2m C Bb21 b22 ... b2p C Bc21 c22 ... c2p C
B C B C B C
B .. .. .. C⋱ B .. .. .. C ¼ B .. .. .. C
@ . . ⋱ . A @ . . ⋱ . A @ . . ⋱ . A
an1 an2 ... anm bn1 bn2 ... bnp cn1 cn2 ... cnp
 2.2 Matrix Algebra 33

where
c11 ¼ a11 b11 þ a12 b21 þ a13 b31 þ . . . a1m bm1
c12 ¼ a11 b12 þ a12 b22 þ a13 b32 þ . . . a1m bm2
..
.
c1p ¼ a11 b1p þ a12 b2p þ a13 b3p þ . . . a1m bmp
c21 ¼ a21 b11 þ a22 b21 þ a23 b31 þ . . . a2m bm1
c22 ¼ a21 b12 þ a22 b22 þ a23 b32 þ . . . a2m bm2
..
.
c2p ¼ a21 b1p þ a22 b2p þ a23 b3p þ . . . a2m bmp
..
.
cn1 ¼ an1 b11 þ an2 b21 þ an3 b31 þ . . . anm bm1
cn2 ¼ an1 b12 þ an2 b22 þ an3 b32 þ . . . anm bm2
..
.
cnp ¼ an1 b1p þ an2 b2p þ an3 b3p þ . . . anm bmp

Matrix multiplication is associative and distributive, but not, in general, commutative:

ðA  BÞ  C ¼ A  ðB  CÞ
A  ðB þ CÞ ¼ A  B þ A  C (2.2.22)
ðA þ BÞ  C ¼ A  C þ B  C
Unless A and B are of the same dimension and diagonal, A  B ≠ B  A.
With regard to the matrix transpose and inverse, the matrix product has the following
properties:

ðA  BÞT ¼ BT  AT
(2.2.23)
ðA  BÞ1 ¼ B1  A1

The determinant of a matrix product is a product of determinants of its multipliers:

jA  B j ¼ jB j  jA j (2.2.24)
Example 2.2.5 Computing the product of two matrices
2 3
2:1 1:0
1:0 2:1 1:0
The product of matrices A ¼ and B ¼ 41:1 2:2 5 is computed by Eq.
3:2 1:1 2:2
(2.2.21) as follows: 0:5 3:0

A  B ¼ C where every element of the product matrix C is

c11 ¼ ð1:0Þ  2:1 þ 2:1  ð1:1Þ þ 1:0  0:5 ¼ 3:91
c12 ¼ ð1:0Þ  1:0 þ 2:1  2:2 þ 1:0  ð3:0Þ ¼ 0:62
c21 ¼ 3:2  2:1 þ ð1:1Þ  ð1:1Þ þ 2:2  0:5 ¼ 9:03
c22 ¼ 3:2  1:0 þ ð1:1Þ  2:2 þ 2:2  ð3:0Þ ¼ 5:82
3:91 0:62
Thus, the product matrix is C ¼
9:03 5:82
34 2 Fundamentals of Mathematics

2.3 Complex Numbers

Im(z)
2.3.1 Definition
A complex number z is given as
jy z
z ¼ x þ jy (2.3.1)
r
where x and y are real numbers,pand
ffiffiffiffiffiffiffi j is the θ Re(z)
imaginary unit of the form j ¼ 1. The x
0 x
and y are the real and imaginary parts of z;
namely,
Figure 2.3.1 Definition of a complex
ReðzÞ ¼ x; ImðzÞ ¼ y (2.3.2)
number

The complex number can be illustrated in a two-dimensional complex plane, as shown in

Figure 2.3.1, where the horizontal and vertical axes are the real and imaginary ones,
respectively. Equation (2.3.1) is an expression of complex numbers in Cartesian coordinates.
From the figure, z can also be expressed in polar coordinates,

z ¼ rðcos θ þ j sin θÞ (2.3.3)

where
x ¼ r cos θ; y ¼ r sin θ (2.3.4)

have been used. Furthermore, by Euler’s formula,

e jθ
¼ cos θ þ j sin θ (2.3.5)

z can be written in the form

z ¼ re jθ (2.3.6)

where θ is in radians. The r and θ in the Eq. (2.3.6) are the magnitude (modulus) and phase
angle (argument) of a complex number z, which by trigonometry formulas are given by
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
jzj ¼ r ¼ x2 þy2
y (2.3.7)
∠ z ¼ θ ¼ tan1
x

Example 2.3.1 Finding the magnitude and phase angle of a complex number
Given a complex number z ¼ 3 þ j4, its magnitude and phase angle are computed by Eq. (2.3.7) as
follows:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
jzj ¼ 32 þ 4 ¼ 5
2

4
∠ z ¼ tan1 ¼ 53:13° ¼ 0:9273 rad
3
 2.3 Complex Numbers 35

The complex number can also be written as

 
z ¼ 5 cosð53:13°Þ þ j sinð53:13°Þ ¼ 5e j0:9273

2.3.2 Phase Angles by Quadrants of a Complex Plane

For complex numbers s1 ¼ 3  j4; s2 ¼
3 þ j4, their phase angles, by Eq. (2.3.7), are Im(z)

given by s1
j4
 
1 4 1 4
θ1 ¼ tan ; θ2 ¼ tan
3 3
θ1
3 Re(z)
Since 4=3 ¼ 4=ð3Þ, it is a common mis-
0
take to conclude that θ1 ¼ θ2 . Because the –3 θ2
arctangent function is multivalued, these
phase angles are not equal: θ1 ≠ θ2 . Hence, we
need to know which quadrants s1 and s2 fall in. –j4 s2
According to Figure 2.3.2, s1 is in the second
quadrant of the complex plane, while s2 is in
the fourth quadrant. Figure 2.3.2 The phase angle

It follows that 
4
θ1 ¼ 180°  tan1 ¼ 126:87°
 3
4
θ2 ¼ tan1 ¼ 53:13°
3

By the same token, one may estimate the following phase angles

1 4
∠ð3  j4Þ ¼ 180° þ tan ¼ 233:13°
3
∠ð3Þ ¼ 0°; ∠ð3Þ ¼ 180°
∠ð j4Þ ¼ 90°; ∠ðj4Þ ¼ 270°

2.3.3 Complex Conjugates

For a complex number z ¼ x þ jy, its complex conjugate is
36 2 Fundamentals of Mathematics

z ¼ x  jy (2.3.8)

where the over-bar stands for complex conjugation. The following properties of the complex
conjugate can be easily derived:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
jzj ¼ jzj ¼ x2 þ y 2
∠ z ¼ ∠ z
z z ¼ jzj 2 ¼ x2 þ y2 (2.3.9)
z þ z ¼ 2x ¼ 2 ReðzÞ
z  z ¼ 2 j ImðzÞ

2.3.4 Algebraic Manipulations

Complex numbers, just like the real ones, can undergo algebraic manipulations such as
addition/subtraction and multiplication/division:

ða þ jbÞ  ðc þ jdÞ ¼ ða  cÞ þ jðb  dÞ

ða þ jbÞ  ðc þ jdÞ ¼ ðac  bdÞ þ jðbc þ adÞ
a þ jb 1 (2.3.10)
¼ 2 fðac þ bdÞ þ jðbc  adÞg
c þ jd c þ d2

Here in the third line of Eq. (2.3.10), the numerator and denominator of the given fraction
were multiplied by the complex conjugate of the denominator, c  jd, and the relation
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
jzj ¼ jzj ¼ x2 þ y2 has been used.

Example 2.3.2 Finding real and imaginary parts of a complex number

3 þ j4
The complex number z ¼ can be represented as a sum of its real and imaginary parts:
4  j7
3 þ j4 3 þ j4 4 þ j7 12  28 þ jð16 þ 21Þ 16 37
z¼ ¼  ¼ ¼ þj
4  j7 4  j7 4 þ j7 4 þ7
2 2 65 65

2.3.5 Trigonometric Manipulations

Since complex numbers can be represented in terms of their magnitudes and phase angles,
they can undergo trigonometric manipulations.
For two complex numbers z1 ¼ r1 e jθ1 and z2 ¼ r2 e jθ2 ,

z1 z2 ¼ re jθ ; with r ¼ r1 r2 ; θ ¼ θ1 þ θ2
z1 r1 (2.3.11)
¼ ρe jβ ; with ρ ¼ ; β ¼ θ1  θ2
z2 r2
 2.4 The Laplace Transform 37

Example 2.3.3 Multiplication of complex numbers using Euler’s formula

Compute the following complex expression

3 þ j4 2  j9
z¼ 
4  j7 10 þ j6

Solution
Write
s1 s2
z¼  ¼ ρe jθ
s3 s4

where by Eq. (2.3.6),

s1 ¼ 3 þ j4 ¼ 5 e j0:9273 ; s2 ¼ 2  j9 ¼ 9:2195 e j4:4937

s3 ¼ 4  j7 ¼ 8:0623 e j1:0517 ; s4 ¼ 10 þ j6 ¼ 11:6619 e j2:6012

It follows from Eq. (2.3.11) that

5 9:2195
ρ¼  ¼ 0:490
8:0623 11:6619
θ ¼ 0:9273 þ 4:4937  ð1:0517Þ  2:6012 ¼ 3:399 rad ¼ 194:7°

Thus,
 
z ¼ 0:490e j3:399 ¼ 0:490 cosð194:7°Þ þ j sinð194:7°Þ ¼ 0:474  j 0:124

2.4 The Laplace Transform

The Laplace transform is a mathematical tool that is widely used for the solution of linear
ordinary differential equations (sometimes abbreviated to “ODEs”) (in particular, linear
time-invariant ordinary differential equations), and for modeling and analysis of dynamic
systems. The Laplace transform is an integral transform, the utility of which lies in convert-
ing linear ordinary differential equations into algebraic equations. It is particularly useful for
handling nonhomogeneous differential equations.
Applications of the Laplace transform are numerous. One of the most common is the
transfer function formulation, which will be discussed in Chapters 3 to 5. Applications of the
Laplace transform to modeling of multi-component dynamic systems, the derivation and
analysis of the system response, and the design and analysis of feedback control systems will
be discussed in Chapters 6 to 8.
38 2 Fundamentals of Mathematics

2.4.1 Definition
For a real-valued, piecewise-continuous function f ðtÞ specified for t ≥ 0, its Laplace trans-
form, denoted by F(s), is defined by
8
< ðT ð∞
st
FðsÞ ≡ L f ðtÞg ¼ limT → ∞ ½f ðtÞ e dt ¼ f ðtÞ est dt (2.4.1)
:
 
0 0

where L is the Laplace transform operator, and s is the complex Laplace transform param-
eter whose real part is greater than some fixed number σ0 . The operational symbol L
indicates that the expression it prefixes is to be transformed by the Laplace Integral
Ð∞ st
e dt. The variable of integration t is arbitrary, and the transform is a function of the
0
complex variable s. While the original function f ðtÞ operates in the time domain, its trans-
form F ðsÞ operates in the Laplace domain.
It needs to be noted that the initial conditions for the Laplace transform are taken at the
time t ¼ 0 to distinguish them from the initial values taken at the time t ¼ 0þ .
When the limit presented by Eq. (2.4.1) is not finite, the Laplace transform for the function
f ðtÞ does not exist. Thus, in order for the Laplace transform of a function f ðtÞ to exist this
limit must be finite, which implies that the Laplace integral must converge. The sufficient
conditions for convergence are:

(a) the function f ðtÞ is piecewise continuous on every finite interval in the range t > 0 and
(b) the function f ðtÞ is of exponential order as t approaches infinity, i.e. it satisfies the
relation j f ðtÞj ≤ Meγt 8t 2 ½0; ∞Þ for some real positive constants M and γ.

Therefore, for the Laplace transform to exist, it is sufficient that

ð∞
j f ðtÞjeσ0 t dt < ∞
0

The reverse process of finding the time function f ðtÞ from its transform F ðsÞ can be
expressed by the inverse Laplace transform:

ð
σþj∞
1 1
f ðtÞ ≡ L fFðsÞg ¼ FðsÞ est ds (2.4.2)
2πj
σj∞

where L1 is the inverse Laplace transform operator, σ p isffiffiffiffiffiffi

called
ffi an abscissa of convergence,
σ > σ0 with σ0 being a real and positive number, and j ¼ 1. The abscissa of convergence is
defined as a real constant and is chosen to be larger than the real parts of all singular points of
 2.4 The Laplace Transform 39

F ðsÞ. This way the path of integration moves to the right of all singular points and is parallel
to the jω axis.
The f ðtÞ and F ðsÞ defined in Eqs. (2.4.1) and (2.4.2) form a Laplace transform pair. Listed
in Table 2.4.1 are the Laplace transform pairs for some basic functions.
In the table, the unit-impulse function (Dirac delta) is defined by

∞; t ¼ 0
δðtÞ ¼ (2.4.3)
0; t ≠ 0

which also satisfies the condition

ða
δðtÞdt ¼ 1 (2.4.4)
b

for any positive a and b. The unit step function is defined by

1; t ≥ 0
uðtÞ ¼ (2.4.5)
0; t < 0

In modeling of dynamic systems, the Dirac delta is used to describe impact inputs, while
the step function represents constant inputs.
The Laplace transform is unique. If two continuous functions f1 ðtÞ and f2 ðtÞ have Laplace
transforms F1 ðsÞ and F2 ðsÞ, respectively, and these transforms are identical, F1 ðsÞ ¼ F2 ðsÞ,
then the functions f1 ðtÞ and f2 ðtÞ are necessarily identical. The unique quality of the Laplace
transform is of great practical significance since it allows one to use it for solving differential
equations by transforming them into algebraic equations in the Laplace domain, solving for
variables in the Laplace domain, and then using the inverse Laplace transform to find the
desired time-domain solution.

2.4.2 Poles and Zeros of the Complex Function F (s)

In many mathematical problems in science and engineering, the Laplace transformed
function F(s) has the following rational form

NðsÞ bm sm þ bm1 sm1 þ . . . þ b1 s þ b0

FðsÞ ¼ ¼ ; m ≤ n; an ≠ 0 (2.4.6)
DðsÞ an sn þ an1 sn1 þ . . . þ a1 s þ a0

where all the coefficients are real. The denominator D(s) has n roots, pi ; i ¼ 1; 2; . . . ; n,
which are called the poles of F(s). In other words, the poles’ pi values satisfy the algebraic
equation
40 2 Fundamentals of Mathematics

Table 2.4.1 Laplace transform pairs

f ðtÞ FðsÞ

Unit impulse function, δðtÞ 1

1
Unit step function, uðtÞ
s
1
Unit ramp function, t
s2
n!
tn
snþ1
1
eat
sþa
n!
tn eat
ðs þ aÞnþ1
ω
sin ωt
s 2 þ ω2
s
cos ωt
s 2 þ ω2
2ω s
t sin ωt
ðs 2 þ ω 2 Þ2
s 2  ω2
t cos ωt
ðs 2 þ ω 2 Þ2
ω
eat sin ωt
ðs þ aÞ2 þ ω2
sþa
eat cos ωt
ðs þ aÞ2 þ ω2
1 1
ðeat  ebt Þ
ba ðs þ aÞðs þ bÞ
1 h i sþp
ðp  aÞeat  ðp  bÞebt
ba ðs þ aÞðs þ bÞ
 qffiffiffiffiffiffiffiffiffiffiffiffiffi
ωn ω2n
pffiffiffiffiffiffiffiffiffiffiffiffiffi eξωn t sin ωn 1  ξ 2 t
1  ξ2 s2 þ 2ξωn s þ ω2n
(  qffiffiffiffiffiffiffiffiffiffiffiffiffi )
 pffiffiffiffiffiffiffiffiffiffiffiffiffi  ξ s
ξωn t
e cos ωn 1  ξ t  pffiffiffiffiffiffiffiffiffiffiffiffiffi sin ωn 1  ξ 2 t
2
1  ξ2 s2 þ 2ξωn s þ ω2n
(  qffiffiffiffiffiffiffiffiffiffiffiffiffi )
 pffiffiffiffiffiffiffiffiffiffiffiffiffi  ξ ω2n
ξωn t  
1e cos ωn 1  ξ t þ pffiffiffiffiffiffiffiffiffiffiffiffiffi sin ωn 1  ξ 2 t
2
1  ξ2 s s2 þ 2ξωn s þ ω2n
 2.4 The Laplace Transform 41

DðsÞ ≡ an sn þ an1 sn1 þ . . . þ a1 s þ a0 ¼ 0 (2.4.7)

The numerator N(s) has m roots, zi ; i ¼ 1; 2; . . . ; m; which are called the zeros of F(s)
and which satisfy the algebraic equation

NðsÞ ≡ bm sm þ bm1 sm1 þ . . . þ b1 s þ b0 ¼ 0 (2.4.8)

It can be shown that

DðsÞ ¼ an ðs  p1 Þðs  p1 Þ . . . ðs  pn Þ
(2.4.9)
NðsÞ ¼ bm ðs  z1 Þðs  z1 Þ . . . ðs  zm Þ

With Eq. (2.4.9), the function F(s) can be expressed in a zero-pole-gain form

ðs  z1 Þðs  z1 Þ . . . ðs  zm Þ an
FðsÞ ¼ k ; k¼ (2.4.10)
ðs  p1 Þðs  p1 Þ . . . ðs  pn Þ bm

Example 2.4.1 Deriving the zero-pole-gain form of a rational function

Consider the rational function

NðsÞ 4s þ 1
FðsÞ ¼ ¼ 2 :
DðsÞ 2s þ 5s þ 3

The poles of F(s) are the roots of

2s2 þ 5s þ 3 ¼ 0 ) p1 ¼ 1:5; p2 ¼ 1

and the zero of F(s) is given by

4s þ 1 ¼ 0 ) z1 ¼ 0:25:

Then the function can be written in the zero-pole-gain form

NðsÞ s þ 0:25
FðsÞ ¼ ¼2 :
DðsÞ ðs þ 1:5Þðs þ 1Þ

2.4.3 Properties of the Laplace Transform

The Laplace transform is a definite integral; thus, it possesses all the properties of such
integrals. The following properties of the Laplace transform are useful in modeling, analysis,
and derivation of analytical solutions for dynamic systems in engineering problems.
42 2 Fundamentals of Mathematics

Factoring

Lfα f ðtÞg ¼ α Lff ðtÞg ¼ α FðsÞ (2.4.11)

This can be proven using the definition of the Laplace transform

ð∞ ð∞
L α f ðtÞg ¼ α f ðtÞest dt ¼ α f ðtÞest dt ¼ α Lff ðtÞg ¼ α FðsÞ
0 0

Superposition (Linearity Property)

Because the Laplace transform operator L is a linear operator, it satisfies the following
superposition principle:

Lfα f1 ðtÞ þ β f2 ðtÞg ¼ α Lff1 ðtÞg þ β Lff2 ðtÞg ¼ α F1 ðsÞ þ β F2 ðsÞ (2.4.12)

The proof is straightforward, using the definition of the Laplace transform and the
properties of definite integrals
Ð∞
L α f1 ðtÞ þ β f2 ðtÞg ¼ 0 ½α f1 ðtÞ þ β f2 ðtÞ est dt
ð∞ ð∞
¼α f1 ðtÞ est dt þ β f2 ðtÞ est dt ¼ α Lff1 ðtÞg þ β Lff2 ðtÞg ¼ α F1 ðsÞ þ β F2 ðsÞ
0 0

The inverse transform also possesses the linearity property

L1 fα F1 ðsÞ þ β F2 ðsÞg ¼ α L 1 fF1 ðsÞg þ β L 1 fF2 ðsÞg

¼ α f1 ðtÞ þ β f2 ðtÞ (2.4.13)

The linearity property is very useful since it allows one to decompose a function into a linear
combination of functions whose transforms are known, thus avoiding the necessity of evalu-
ating the Laplace integral.

Example 2.4.2 Evaluating the Laplace transform of a function using the table
of Laplace pairs
  
L e4t ð5 sin 2t þ 3tÞ ¼ 5L e4t sin 2t þ 3L te4t
10 3
¼ 2
þ
ðs þ 4Þ þ 2 2
ðs þ 4Þ2

where Table 2.4.1 has been used to evaluate the Laplace transform of the given functions.
 2.4 The Laplace Transform 43

Example 2.4.3 Deriving the Laplace transform of the common functions: step, ramp,
exponential, and sinusoidal
Deriving the Laplace transform of some common functions.
Step Function
The step function is defined as

0 t<0
f ðtÞ ¼
A t≥0

where A is a real positive constant, called the magnitude of the step function. If A ¼ 1, the function
is called the unit step function, and it is often denoted as uðtÞ. Hence, any step function can be
expressed as

0 t<0
f ðtÞ ¼
A uðtÞ t ≥ 0

Using the definition and the above-described properties, the Laplace transform of a step
function is
ð∞
∞
A
A A
Lf AuðtÞg ¼ AuðtÞest dt ¼  est
 ¼  ðe ∞  e0 Þ ¼
0 s 0 s s

Obviously, the Laplace transform of the unit step is as shown in Table 2.4.1
ð∞
∞
1
1 1
Lf uðtÞg ¼ uðtÞest dt ¼  est
¼  ðe ∞  e0 Þ ¼
0 s 0 s s

Ramp Function
The ramp function is defined as follows:
0 t<0
f ðt Þ ¼ , where A is a real positive constant. If A ¼ 1, the function is called unit ramp
A t t≥0
function.
The Laplace transform of a ramp function is evaluated using the technique of integration by
parts:
ð∞ ð∞  ð
st A st

∞ A st A ∞ st A
L f A tg ¼ Ate dt ¼  t e
  e dt ¼ e dt ¼ 2
0 s 0 0 s s 0 s

Exponential Function
The exponential function is defined as follows:
0 t<0
f ðt Þ ¼ , where A and α are real positive constants.
A eαt t ≥ 0
The Laplace transform of this function is
44 2 Fundamentals of Mathematics

ð∞ ð∞
A
Lf A eαt g ¼ A eαt est dt ¼ A eðαþsÞt dt ¼
0
0 sþα

Sinusoidal Function
The sinusoidal function is defined as follows:

0 t<0
f ðtÞ ¼
A sin ωt t ≥ 0

where A and ω are real positive constants.

Remembering the Euler’s formula, ejθ ¼ cos θ þ j sin θ and ejθ ¼ cos θ  j sin θ, the Laplace
transform of the sine and cosine function are derived as follows. From the Euler’s formula:

1 jω t
sin ωt ¼ ðe  ejωt Þ
2j
1
cos ωt ¼ ðe jω t þ ejωt Þ
2

Hence, the Laplace transform of the sine function is

ð∞
A A 1 A 1 Aω
Lf A sin ωtg ¼ ðe jωt  ejωt Þest dt ¼    ¼
2j 0  2j s  jω 2j s þ jω s2 þ ω2

and for the cosine function

0 t<0
f ðtÞ ¼
A cos ωt t ≥ 0

where A and ω are real positive constants, the Laplace transform is

ð∞
A A 1 A 1 As
Lf A cos ωtg ¼ ðe jωt þ ejω t Þest dt ¼  þ  ¼
2 0 2 s  jω 2 s þ jω s2 þ ω2

Time Delay (Shifting along the Time Axis)

For a function f ðtÞ, define a time-shifted function fd ðtÞ by

f ðt  TÞ; t ≥ T
fd ðt  TÞ ¼ (2.4.14)
0; 0 ≤ t < T

where T is a real positive constant, called time-delay parameter.

The time-shifted function is illustrated in Figure 2.4.1.
By definition, the Laplace transform of the translated function fd ðt  TÞ is computed as
 2.4 The Laplace Transform 45

f (t )
fd (t – T )

f (t )

0 T
t

Figure 2.4.1 Time-shifted function

ð∞
Lffd ðt  T Þg ¼ f ðt  T Þest dt
0

Let the independent variable be defined as τ ¼ t  T. Then the Laplace integral becomes
ð∞ ð∞
st
f ðt  TÞe dt ¼ f ðτÞesðτþTÞ dτ
0 T

Since f ðt  T Þ ¼ f ðτÞ ¼ 0 for τ < 0, the lower limit of the integration can be changed from
T to 0. Hence, the Laplace integral is evaluated as
ð∞ ð∞
Ð∞ sðτþTÞ sτ sT sT
T f ðτ Þe dτ ¼ f ðτÞe e dτ ¼ e f ðτÞesτ dτ
0 0

¼ esT L f ðtÞg ¼ esT FðsÞ

Therefore, translation of the function f ðtÞ in time domain by T ≥ 0 corresponds to the

multiplication of its Laplace transform F ðsÞ by esT

L fd ðt  TÞg ¼ esT FðsÞ (2.4.15)

Example 2.4.4 Deriving the Laplace transform of a time-shifted function

For a time-shifted step function

1; t ≥ T
uðt  TÞ ¼
0; t < T

where T > 0, the Laplace transform is

1
Lfuðt  TÞg ¼ esT LfuðtÞg ¼ esT :
s
46 2 Fundamentals of Mathematics

Multiplication by the Exponential (Shifting along the s-Axis)

Similarly to translation in the time domain, a function may be translated in the Laplace
domain. If f ðtÞ has the Laplace transform F ðsÞ, then the Laplace transform of a function
eαt f ðtÞ will be
ð∞ ð∞
αt αt st
Lf e f ðtÞg ¼ e f ðtÞe dt ¼ f ðtÞeðαþsÞt dt ¼ Fðs þ αÞ (2.4.16)
0 0

Multiplication by t
Let F ðsÞ be the Laplace transform of the function f ðtÞ. Differentiating F ðsÞ once with respect
to s, we obtain
ð∞ ð∞
d d
FðsÞ ¼ f ðtÞest dt ¼  t f ðtÞest dt ¼ Lft f ðtÞg
ds ds 0 0

Hence
d
Lft f ðtÞg ¼  FðsÞ (2.4.17)
ds

Extending this approach to the powers of t, the following generic expression is derived

dn
Lf tn f ðtÞg ¼ ð1Þn FðsÞ (2.4.18)
dsn

Example 2.4.5 Deriving the Laplace transform of various functions using the table
of Laplace transform Pairs and Laplace transform properties
Certain functions need to be evaluated to facilitate derivation of their Laplace transform. For
example, a function f ðtÞ ¼ ða þ ktÞ2 needs to be converted to a polynomial before its Laplace
transform can be derived:

Lðf ðtÞÞ ¼ Lðða þ ktÞ2 Þ ¼ Lða2 þ 2akt þ k2 t2 Þ

The linearity property is used:

Lðf ðtÞÞ ¼ Lða2 þ 2akt þ k2 t2 Þ ¼ Lða2 Þ þ Lð2aktÞ þ Lðk2 t2 Þ

Considering that a2 represents a step function a2 uðtÞ, apply a factoring property and then use
Table 2.4.1 to derive the solution:

a2 2ak 2k2
Lðf ðtÞÞ ¼ Lða2 Þ þ 2akLðtÞ þ k2 Lðt2 Þ ¼ þ 2 þ 3
s s s
 2.4 The Laplace Transform 47

Trigonometric identity needs to be used to convert the following function to a form that allows the
use of Table 2.4.1 to derive the required Laplace transform:

1
Lðf ðtÞÞ ¼ Lðsin2 ðωtÞÞ ¼ L ð1  cosð2ωtÞ
2
1 1 1 s
¼ LðuðtÞÞ  Lðcosð2ωtÞÞ ¼ 
2 2 2s 2ðs2 þ 4ω2 Þ

When several properties of the Laplace transform need to be used to derive the transform of a
given function, linearity and factoring properties are used first, followed by the other properties as
needed.
In some cases, the sequence of applications of the Laplace transform properties is uniquely
defined. For example, while proving that
  s2  ω2
L t cosðωtÞ ¼
ðs2 þ ω2 Þ2

the transform of a cosine needs to be evaluated first, followed by application of the multiplication
by t property: 
  d s s2  ω2
L t cosðωtÞ ¼  ¼
ds s þ ω
2 2
ðs2 þ ω2 Þ2

In other cases, the Laplace transform properties could be applied in a different order. For
example, the function f ðtÞ ¼ 3t sinhð2tÞ needs first to be represented as a sum of products of
exponential functions and t, and then the appropriate Laplace transform properties applied as
follows: 
1
Lð f ðtÞÞ ¼ Lð3t sinhð2tÞÞ ¼ L ð3te2t  3te2t Þ
2

Linearity and factoring properties are applied first:


1 3 3
Lð f ðtÞÞ ¼ L ð3te2t  3te2t Þ ¼ Lðte2t Þ  Lðte2t Þ
2 2 2

Then two options are available:

• Use the Laplace transform of an exponential function and the multiplication by t property
  
3 3 2t 3 d 1 d 1
Lð f ðtÞÞ ¼ Lðte Þ  Lðte Þ ¼
2t
 þ
2 2 ! 2 ds s  2 ds s þ 2
3 1 1 12s
¼  ¼
2 ðs  2Þ2 ðs þ 2Þ2 ðs  4Þ2
2
48 2 Fundamentals of Mathematics

• Use Laplace transform of a unit ramp function and the multiplication by an exponential (shifting
along the s-axis) property
!
3 3 2t 3 1 1 12s
Lðf ðtÞÞ ¼ Lðte Þ  Lðte Þ ¼
2t
2
 2
¼
2 2 2 ðs  2Þ ðs þ 2Þ ðs  4Þ2
2

Differentiation (Derivative Property)

The Laplace transform of the first derivative of a function f ðtÞ is obtained by using integra-
tion by parts to evaluate the Laplace integral:
 ð∞
∞ ð∞
d df st

L f ðtÞ ¼ e dt ¼ f ðtÞest
 þ s f ðtÞest dt
dt 0 dt 0 0

¼ f ð0 Þ þ sLff ðtÞg ¼ sFðsÞ  f ð0 Þ (2.4.19)

Similarly, the Laplace transform of the second derivative of a function f ðtÞ is


d2 d
L f ðtÞ ¼ s2 FðsÞ  s f ð0 Þ  f ð0 Þ (2.4.20)
dt2 dt

Extending the above procedure to the higher derivatives, the general case expression is
obtained:

dn f ðtÞ
L ¼ sn FðsÞ  sn1 f ð0 Þ  sn2 f ð1Þ ð0 Þ  . . .  f ðn1Þ ð0 Þ (2.4.21)
dtn
dk f
where f ðkÞ ¼ .
dtk
Equation (2.4.21) can be written in a more compact form
 Xn
dn f ðtÞ
L ¼ s n
FðsÞ  snk gk1
dtn k¼1

dk1 f

P
where gk1 ¼ k1
 and; obviously; the sum nk¼1 snk gk1 represents n initial condi-
dt t¼0
tions, defined for the n-th order differential equation.

Example 2.4.6 Deriving the Laplace transform of a function derivative

d 4t
Given derivative φðtÞ ¼ e ð5sin 2t þ 3tÞg, its Laplace transform is
dt
 10s 3s
LfφðtÞg ¼ s L e4t ð5sin 2t þ 3tÞ  ½e4t ð5sin 2t þ 3tÞt¼0 ¼ 2
þ
ðs þ 4Þ þ 2 2
ðs þ 4Þ2

where the result derived in the Example 2.4.2 has been used.
 2.4 The Laplace Transform 49

The utility of the derivative property of the Laplace transform (Eq. (2.4.21)) and its
linearity property (Eq. (2.4.12)) lies in the ability to convert a differential equation in the
time domain into an algebraic equation in the Laplace domain. This significantly facilitates
finding the solution of a given differential equation. Such an approach is illustrated in the
Example 2.4.7.

Example 2.4.7 Solving a linear differential equation with a Laplace transform

Consider the following linear second-order differential equation with initial conditions

_ þ kxðtÞ ¼ f ðtÞ
x ðtÞ þ cxðtÞ
m€
xð0 Þ ¼ x0 ; xð0
_  Þ ¼ v0

dx
where x_ ¼ ; and m; c; and k are constants. The previous equation and initial conditions are
dt
often used to describe a spring–mass–damper system, which is to be discussed in Chapter 3.
According to Eqs. (2.4.12) and (2.4.21), the Laplace transform of this differential equation yields
   
m s2 X ðsÞ  sx0  v0 þ c sX ðsÞ  x0 þ kX ðsÞ ¼ FðsÞ

which is reduced to an algebraic equation in terms of X ðsÞ ¼ LfxðtÞg:

 
ms2 þ cs þ k X ðsÞ ¼ FðsÞ þ mx0 s þ mv0 þ cx0

Thus, the s-domain (Laplace transform domain) solution of the given differential equation is

FðsÞ þ mx0 s þ mv0 þ cx0

X ðsÞ ¼
ms2 þ cs þ k

The inverse Laplace transform of the previous expression gives the time-domain solution of the
original differential equation; that is xðtÞ ¼ L 1 fX ðsÞg.
This example clearly demonstrates the procedure of using a Laplace transform to solve linear
ordinary differential equations with constant coefficients. This technique shall be investigated
further in Section 2.6.3.

Integration (Integral Property)

The Laplace transform of an integral of a function f ðtÞ is derived using integration by parts
and defining a function e
f ð0 Þ such that
ð

e

f ð0 Þ ¼ f ðtÞdt

t¼0
50 2 Fundamentals of Mathematics

Ð 
Let L f ðtÞdt ¼ FðsÞ. Then the Laplace transform of an indefinite integral of the
function f ðtÞ is
ð ∞ ð ð ð∞
Ð  est

∞ est
L f ðtÞdt ¼ f ðtÞdt est dt ¼ f ðtÞdt
 f ðtÞ dt
0 s 0 0 s
ð
ð
1
1 ∞ 1 1
¼ f ðtÞdt
 þ f ðtÞdt ¼ ef ð0  Þ þ FðsÞ (2.4.22)
s t¼0 s 0 s s

Since
ðt ð
f ðtÞdt ¼ f ðtÞdt  e
f ð0  Þ
0

The application of the result of Eq. (2.4.22) to the definite integral of f ðtÞ derives the integral
property as follows:
ð∞  ð  n o
L f ðtÞdt ¼ L f ðtÞdt  L e f ð0  Þ (2.4.23)
0

Since e
f ð0  Þ is a constant, its Laplace transform is derived using the linearity property and
the known Laplace transform of a unit step function
n o n o 
1
e  e  e
L f ð0 Þ ¼ L f ð0 ÞuðtÞ ¼ f ð0 Þ L uðtÞ ¼ e

f ð0  Þ
s

Applying this result to the Eq. (2.4.23), we obtain the sought integral property
ð∞ 
1 1 1 1
L f ðtÞdt ¼ e f ð0  Þ þ FðsÞ  e f ð0  Þ ¼ FðsÞ (2.4.24)
0 s s s s

Example 2.4.8 Deriving the Laplace transform of a linear integro-differential

expression
Consider the following linear integro-differential expression

ðt
deðtÞ
uðtÞ ¼ kp eðtÞ þ ki eðτÞdτ þ kd
dt
0

where the function eðtÞ is specified for t ≥ 0, and kp ; ki ; kd are constants. Assuming a zero initial
value of eðtÞ, namely eð0Þ ¼ 0, the Laplace transform of the above expression gives

1
UðsÞ ¼ kp þ ki þ kd s EðsÞ
s

This s-domain expression describes an algorithm widely used in control engineering, called the
PID feedback control law.
 2.4 The Laplace Transform 51

Initial-Value Theorem
The initial-value theorem allows the use of the Laplace transform of a function f ðtÞ to find
the function value at a time slightly greater than zero, namely f ð0þ Þ. It becomes important in
the analysis of dynamic systems under impulse input.
A unit impulse function (Dirac delta), as defined in Eq. (2.4.3), is a mathematical approxi-
mation of an infinitely large input of energy into a system, which occurs within an infinitesi-
mally short time.
If the Laplace transform of a given function f ðtÞ and its first derivative with respect to time
df
exist, and lim sFðsÞ also exists, the initial-value theorem is formulated as
dt s→0

f ð0þ Þ ¼ limþ f ðtÞ ¼ lim sFðsÞ (2.4.25)

t→0 s→∞

Example 2.4.9 Finding the system response to a unit impulse using the Laplace
transform and the initial-value theorem
A dynamic system is described by a second-order differential equation with initial
conditions
x€ ¼ f ðtÞ
xð0 Þ ¼ 5; xð0
_  Þ ¼ 10

Find the system response to a unit impulse, namely values of system parameters xðtÞ and x_ ðtÞ at
t ¼ 0þ when f ðtÞ ¼ δðtÞ.
Solution
Since the Dirac delta function is defined for t ¼ 0 only, and equals zero for all other values of t, it is
logical that the initial conditions are given at a time t ¼ 0  that immediately precedes t ¼ 0, when
the impulse input was applied.
Taking the Laplace transform of the given differential equation and considering that
X ðsÞ ¼ LfxðtÞg, we obtain

s2 X ðsÞ  sxð0  Þ  xð0

_  Þ ¼ s2 X ðsÞ  5s  10 ¼ 1
5s þ 11
X ðsÞ ¼
s2

Using the initial-value theorem, compute


5s þ 11 11
xð0 þ Þ ¼ lim sX ðsÞ ¼ lim s ¼ lim 5 þ lim ¼5
s→∞ s→∞ s2 s→∞ s→∞ s
52 2 Fundamentals of Mathematics

Recalling that
 
xð0þ Þ ¼ lim sX ðsÞ ¼ lim sLfxðtÞg
s→∞ s→∞

derive
    
_ þ Þ ¼ lim
xð0 _
sLfxðtÞg ¼ lim s sX ðsÞ  xð0 Þ
s→∞ s→∞

 
5s þ 11
¼ lim s s 5 ¼ 11
s→∞ s2

Final-Value Theorem
If a function f ðtÞ converges to some constant value after a long time, namely t → ∞, it is said
that f ðtÞ attains a steady state and that limiting constant is called the steady-state value. A
counterpart of the initial-value theorem, the final-value theorem allows the Laplace trans-
form of f ðtÞ to be used for finding its steady-state value, lim f ðtÞ.
t→∞
Let f ðtÞ and F ðsÞ be a Laplace transform pair. If sF ðsÞ is analytic on the imaginary axis and
the right half of the s-plane (namely, all the poles of sF ðsÞ have negative real parts), then

fss ¼ lim f ðtÞ ¼ lim sFðsÞ (2.4.26)

t→∞ s→0

Here, the poles of a complex function have been defined in Eq. (2.4.7).

Convolution Theorem
If F1 ðsÞ ¼ Lff1 ðtÞg and F2 ðsÞ ¼ Lff2 ðtÞg, then
ðt
L 1 fF1 ðsÞF2 ðsÞg ¼ f1  f2 ≡ f1 ðτÞf2 ðt  τÞdτ (2.4.27)
0

where f1 ðt  τÞ is time-shifted in relation to f1 ðtÞ, and f1  f2 is called the convolution integral of

f1 and f2 . The convolution integral is commutative, namely
ðt ðt
f 1  f2 ≡ f1 ðτÞf2 ðt  τÞdτ ¼ f2  f1 ≡ f2 ðτÞf1 ðt  τÞdτ (2.4.28)

0 0

where f2 ðt  τÞ is time-shifted in relation to f2 ðtÞ.

 2.4 The Laplace Transform 53

The utility of the final-value theorem and the convolution theorem will be shown in
Chapter 7.

2.4.4 Inverse Laplace Transform via Partial Fraction Expansion

In engineering analysis, the inverse Laplace transform is frequently performed to derive
analytical solutions. As mentioned in Example 2.4.7, the inverse Laplace transform of
X(s) can yield the solution of the differential equation. Although extensive Laplace-
transform tables like Table 2.4.1 are available, they may not be sufficient for finding
the inverse Laplace transform of rational functions. Quite often a technique called
partial fraction expansion (or partial fraction decomposition) is used. Partial fraction
expansion has been introduced in integral calculus. This technique is briefly reviewed
below.
The utility of partial fraction decomposition lies in the ability of expanding F ðsÞ into
α αs þ β
a linear combination of components such as and 2 , the inverse
sþp s þ as þ b
Laplace transforms of which are available in any table of Laplace transform pairs. From
Table 2.4.1

1 α
L ¼ αept
sþp

1 αs þ β
L ¼ eσt ðC1 cos ωt þ C2 sin ωtÞ
s2 þ as þ b

where
a
σ¼
2
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4b  a2
ω¼
2
2β  aα
C1 ¼ α and C2 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4b  a2

Consider a rational function F(s) of the form

NðsÞ bm sm þ bm1 sm1 þ . . . þ b1 s þ b0

FðsÞ ¼ ¼ n ; m<n (2.4.29)
DðsÞ s þ an1 sn1 þ . . . þ a1 s þ a0

Here, for simplicity in discussion, the unity coefficient for the sn term and m < n have been
assumed. If n = m, long division can be applied to obtain
54 2 Fundamentals of Mathematics

N1 ðsÞ
FðsÞ ¼ γ þ (2.4.30)
DðsÞ

where γ is a constant, and the order of polynomial N1 ðsÞ is less than n. In this case a partial
fraction expansion can be performed on N1 ðsÞ=DðsÞ: Because of this the subsequent discus-
sion will be focused on Eq. (2.4.29), although the results are also applicable to Eq. (2.4.30).
Let the poles of F ðsÞ be pi ; i ¼ 1; 2; . . . ; n. These poles can be real and complex. In what
follows, the partial fraction expansion and inverse Laplace transform are presented for the
two cases of real poles of F(s).

Case 1: All Poles of F(s) are Real and Distinct

Assume that no two poles of F(s) are the same. Then the partial fraction expansion of F(s) is
r1 r2 rn
FðsÞ ¼ þ þ ... þ (2.4.31)
s  p1 s  p2 s  pn

where ri is the residue of F(s) associated with the pole pi and it is given by

NðsÞ
ri ¼ lim ðs  pi Þ (2.4.32)
s → pi DðsÞ

Instead of computing the above limit the residues can be determined using the cover-up
method, which performs cross-multiplication as follows:

NðsÞ ¼ FðsÞDðsÞ
 (2.4.33)
r1 r2 rn
¼ þ þ ... þ ðs  p1 Þðs  p2 Þ . . . ðs  pn Þ
s  p1 s  p2 s  pn

This eventually leads to

bm sm þ bm1 sm1 þ . . . þ b1 s þ b0 ¼ r1 ðs  p2 Þðs  p3 Þ . . . ðs  pn Þ
þ r2 ðs  p1 Þðs  p3 Þ . . . ðs  pn Þ (2.4.34)
þ . . . þ rn ðs  p1 Þðs  p2 Þ . . . ðs  pn1 Þ

Comparison of the coefficients of the s power terms on the both sides of Eq. (2.4.34)
gives coupled algebraic equations in terms of residues, which are then solved for
these residues. Upon computing the residues, the inverse Laplace transform of F(s) is
found as
 Xn
1
f ðtÞ ¼ L FðsÞ ¼ r1 ep1 t þ r2 ep2 t þ . . . þ rn epn t ¼ r epi
i¼1 i
t
(2.4.35)

by Eq. (2.4.29) and the fifth entry of Table 2.4.1.

 2.4 The Laplace Transform 55

Example 2.4.10 Using partial fraction expansion to derive the inverse Laplace
transform of a given rational function
Consider the function F(s) in Example 2.4.1:

NðsÞ 4s þ 1 2s þ 0:5
FðsÞ ¼ ¼ ¼
DðsÞ 2s2 þ 5s þ 3 s2 þ 2:5s þ 1:5

As previously found, the poles of F ðsÞ are p1 ¼ 1:5 and p2 ¼ 1. The partial fraction
expansion of the function F ðsÞ is then

2s þ 0:5 r1 r2
FðsÞ ¼ ¼ þ :
ðs þ 1:5Þðs þ 1Þ s þ 1:5 s þ 1

According to Eq. (2.4.32) residues r1 and r2 are computed as


r1 ¼ lim 2s þ 0:5 3 þ 0:5
ðs þ 1:5Þ ¼ ¼5
s → 1:5 ðs þ 1:5Þðs þ 1Þ 0:5

2s þ 0:5 2 þ 0:5
r2 ¼ lim ðs þ 1Þ ¼ ¼ 3
s → 1 ðs þ 1:5Þðs þ 1Þ 0:5

Subsequently

5 3
f ðtÞ ¼ L 1  ¼ 5e1:5t  3et :
s þ 1:5 s þ 1

On the other hand, the residues can be found by the cover-up method,

2s þ 0:5 ¼ ðs þ 1:5Þðs þ 1ÞFðsÞ ¼ r1 ðs þ 1Þ þ r2 ðs þ 1:5Þ

which, through comparison of s power terms, yields

s1 terms : 2 ¼ r1 þ r2
s0 terms : 0:5 ¼ r1 þ 1:5r2

The solution of the above algebraic equations gives the same residues as the previous method:
r1 ¼ 5 and r2 ¼ 3, which leads to the same partial fraction decomposition and the same result
of the inverse Laplace transform.

Example 2.4.11 Using long division followed by partial fraction expansion to derive
the inverse Laplace transform of a given function
Consider the function

NðsÞ 3s2 þ 2s þ 0:5

FðsÞ ¼ ¼
DðsÞ s2 þ 2:5s þ 1:5
56 2 Fundamentals of Mathematics

which has the same denominator as the function in Example 2.4.10, but features an added second-
order term 3s2 in the numerator. By long division

5:5s þ 4
FðsÞ ¼ 3  F1 ðsÞ; with F1 ðsÞ ¼
s2 þ 2:5s þ 1:5

Now perform partial fraction expansion on F1 ðsÞ

5:5s þ 4 r1 r2
F1 ðsÞ ¼ ¼ þ
ðs þ 1:5Þðs þ 1Þ s þ 1:5 s þ 1

where

r1 ¼ lim 5:5s þ 4
ðs þ 1:5Þ ¼ 8:5
s → 1:5 ðs þ 1:5Þðs þ 1Þ

5:5s þ 4
r2 ¼ lim ðs þ 1Þ ¼ 3
s → 1 ðs þ 1:5Þðs þ 1Þ

Thus,
8:5 3
FðsÞ ¼ 3  þ
s þ 1:5 s þ 1

and its inverse Laplace transform is

f ðtÞ ¼ 3δðtÞ  8:5e1:5t þ 3et

where Table 2.4.1 has been used, and δðtÞ is the Dirac delta function defined in Eqs. (2.4.3) and
(2.4.4).

Case 2: Poles of F(s) are Real and Repeated

Assume that F(s) has q distinct and k identical poles:

DðsÞ ¼ ðs  p1 Þðs  p2 Þ . . . ðs  pq Þðs  p0 Þk (2.4.36)

where k þ q ¼ n, and poles p0 ; p1 ; p2 . . . pq are distinct. The integer k is the multiplicity of

the pole p0 . In this case, the partial fraction expansion of F(s) takes the form
r1 r2 rq
FðsÞ ¼ þ þ ... þ
s  p1 s  p2 s  pq
(2.4.37)
b1 b2 bk
þ þ þ ... þ
ðs  p0 Þ k
ðs  p0 Þk1 s  p0
 2.4 The Laplace Transform 57

where on the right-hand side the first q terms are about the distinct poles p1 ; p2 . . . pq , and
the remaining k terms are related to the k repeated poles (p0 ). The r1 ; r2 . . . rq are the
residues associated with poles p1 ; p2 . . . pq , while b1 ; b2 . . . bk are the residues associated
with the pole p0 . The formulae for evaluating the residues are given below:

NðsÞ
ri ¼ lim ðs  pi Þ ; i ¼ 1; 2; . . . ; q (2.4.38)
s → pi DðsÞ
1 di1 ΔðsÞ
bi ¼ lim ; i ¼ 1; . . . ; k (2.4.39)
ði  1Þ! s → p0 dsi1

where i is the “running number,” and

NðsÞ
ΔðsÞ ¼ FðsÞðs  p0 Þk ¼ (2.4.40)
ðs  p1 Þðs  p2 Þ . . . ðs  pq Þ

with N(s) given in Eq. (2.4.29). Using Table 2.4.1 and Eq. (2.4.32), the inverse Laplace
transform of F(s) is found as

f ðtÞ ¼ r1 ep1 t þ r2 ep2 t þ . . . þ rq epq t þ


b1 b2 bk1 bk (2.4.41)
þ ep0 t tk1 þ tk2 þ . . . þ tþ t
ðk  1Þ! ðk  2Þ! 1! 0!

Example 2.4.12 Using partial fraction expansion to derive the inverse Laplace
transform of a given rational function with repeated poles
The function

5ðs þ 2Þ
FðsÞ ¼
s2 ðs þ 1Þðs þ 3Þ

has poles 0; 0;  1;  3, with 0 being a pole of multiplicity 2 (k ¼ 2). The partial fraction
expansion of F(s) can be written as

r1 r2 b1 b2
FðsÞ ¼ þ þ 2þ
sþ1 sþ3 s s

The residues associated with 1 and  3 are

  5
r1 ¼ lim ðs þ 1ÞFðsÞ ¼
s → 1 2
  5
r2 ¼ lim ðs þ 3ÞFðsÞ ¼
s → 3 18
58 2 Fundamentals of Mathematics

For the residues associated with the repeated poles, consider the function
5ðs þ 2Þ
ΔðsÞ ¼ s2 FðsÞ ¼
ðs þ 1Þðs þ 3Þ

By Eq. (2.4.39), residues associated with the repeated poles are

1 10
b1 ¼ lim ΔðsÞ ¼
0! s → 0 3
!
1 dΔðsÞ ðs þ 1Þðs þ 3Þ  ðs þ 2Þð2s þ 4Þ 25
b2 ¼ lim ¼ lim 5 ¼
1! s → 0 ds s→0 ðs þ 1Þ2 ðs þ 3Þ2 9

Thus,

1 r1 r2 b1 b2 5 5 10 25
f ðtÞ ¼ L þ þ 2þ ¼ et þ e3t þ t 
sþ1 sþ3 s s 2 18 3 9

2.4.5 Imaginary and Complex Poles in the Inverse Laplace Transform

In many engineering problems the poles and residues of a rational function F ðsÞ are complex
and/or imaginary, which yields complex terms in Eqs. (2.4.35) and (2.4.41). Because F ðsÞ in
Eq. (2.4.29) only contains real coefficients (ai and bi ), its imaginary and complex poles always
pffiffiffiffiffiffiffi
come in pairs, such as jω and α  jβ, where j ¼ 1. As a result, the inverse Laplace
transform of F ðsÞ is real-valued. In what follows, we derive real expressions for the inverse
Laplace transform of F(s) with imaginary or complex poles.
If p is a complex pole and r is the residue of F ðsÞ at this pole, the terms r ept þ r ept will be
included in Eq. (2.4.35), where p and r are the complex conjugates of p and r, respectively.
pffiffiffiffiffiffiffi
Let p ¼ σ þ jω; r ¼ α þ jβ; with j ¼ 1. It follows that
 
r ept þ r ept ¼ 2Re ðα þ jβÞ eðσþjωÞt ¼ 2eσt ðα cos ωt  β sin ωtÞ (2.4.42)

If p is an imaginary pole, say, p ¼ jω, Eq. (2.4.35) will contain the terms r ept þ r ept . By
writing r ¼ α þ jβ,
 
r ept þ r ept ¼ 2Re ðα þ jβÞ ejωt ¼ 2ðα cos ωt  β sin ωtÞ (2.4.43)

Note that Eq. (2.4.43) is a special case of Eq. (2.4.42) with σ ¼ 0.

Another way to deal with complex/imaginary poles is to keep quadratic forms in a partial
fraction expansion, such as
 2.4 The Laplace Transform 59

as þ b as þ b
with c2 < 4d and 2
s2 þ cs þ d s þ ω2

which have complex and imaginary poles. By proper scaling and through use of Table 2.4.1,
the real expression of L 1 fFðsÞg can be directly obtained without having to deal with
complex quantities.

Example 2.4.13 Using partial fraction expansion to derive the inverse Laplace
transform of a given rational function with complex poles
We would like to solve the differential equation in Example 2.4.7, with parameters
_
m ¼ 1; c ¼ 2; k ¼ 2, initial conditions xð0Þ ¼ 1 and xð0Þ ¼ 0; and the forcing function being a
unit step function, f ðtÞ ¼ uðtÞ. According to the result derived in Example 2.4.7, the s-domain
solution of this differential equation is

1 1 s2 þ 2s þ 1
X ðsÞ ¼ þ s þ 2 ¼
s2 þ 2s þ 2 s sðs2 þ 2s þ 2Þ

The poles of X ðsÞ are p1 ¼ 0; p2 ¼ 1 þ j; and p3 ¼ p 2 ¼ 1  j: Because the poles are distinct,
the inverse Laplace transform of X ðsÞ is of the form presented for Case 1:

xðtÞ ¼ r1 ep1 t þ r2 ep2 t þ r3 ep3 t

where the residues by Eq. (2.4.32) are obtained as

1 1 1
r1 ¼ ; r2 ¼ ð1  jÞ; r3 ¼ r 2 ¼ ð1 þ jÞ
2 4 4

According to Eq. (2.4.42),


1 1 1 1 t
xðtÞ ¼ þ 2Re ð1  jÞ eð1þjÞt ¼ þ e ðcos t þ sin tÞ:
2 4 2 2

As another method for the inverse Laplace transform, write

r as þ b
X ðsÞ ¼ þ 2
s s þ 2s þ 2

where r, a, and b are coefficients to be determined. With the method shown in Eq. (2.4.34), we
obtain
 
s2 þ 2s þ 1 ¼ s s2 þ 2s þ 2 X ðsÞ

  r as þ b
¼ s s2 þ 2s þ 2 þ 2
s s þ 2s þ 2
2 
¼ r s þ 2s þ 2 þ sðas þ bÞ
60 2 Fundamentals of Mathematics

Comparison of the coefficients of the s power terms on the two sides of the previous equation
yields

s2 : 1 ¼ r þ a
s1 : 2 ¼ 2r þ b
s0 : 1 ¼ 2r

The solution of the above coupled algebraic equations gives

1 1
r¼ ; a¼ ; b¼1
2 2

Hence,
   !
1 1 sþ2 1 1 sþ2
X ðsÞ ¼ þ ¼ þ
2s 2 s2 þ 2s þ 2 2s 2 ðs þ 1Þ2 þ 1
 !  !
1 1 sþ1 1 1
¼ þ þ
2s 2 ðs þ 1Þ2 þ 1 2 ðs þ 1Þ2 þ 1

where the expression for X ðsÞ is properly scaled and arranged for utility of Table 2.4.1. It follows
from the Laplace transform table that the solution of the differential equation is

1 1 t
xðtÞ ¼ þ e ðcos t þ sin tÞ
2 2

which is the same as obtained previously.

Examples 2.4.7 and 2.4.13 indicate a method for solution of linear differential equations with
constant coefficients, which will be further explored later on in Section 2.6.

2.4.6 Piecewise-Continuous Functions in the Laplace Transformation

In many engineering applications, dynamic systems are subject to piecewise-continuous
inputs such as, for example, pulses shown in Figure 2.4.2. These pulses physically represent
external excitations persisting for the finite time duration.
The Laplace transform of this type of function can be performed through the application
of the principle of superposition as described by Eq. (2.4.12) and the formula for time-shifted
functions as given in Eq. (2.4.15). In general, a piecewise-continuous function can be
decomposed as

f ðtÞ ¼ f0 ðtÞ uðtÞ þ f1 ðt  T1 Þ uðt  T1 Þ þ f2 ðt  T2 Þ uðt  T2 Þ þ . . . (2.4.44)

 2.4 The Laplace Transform 61

f(t) q(t )

q0

F0
α
1
0 0
t t
T T
(a) (b)

Figure 2.4.2 Piecewise-continuous functions: (a) rectangular pulse f(t), (b) triangular pulse q(t)

where f0 ðtÞ; f1 ðtÞ; f2 ðtÞ; . . . are basic functions, as those in Table 2.4.1, f1 ðt  T1 Þ;
f2 ðt  T2 Þ; . . . are time-shifted functions, and T1 ; T2 ; . . . are positive time-delay param-
eters. The Laplace transform of f ðtÞ then has the form

FðsÞ ¼ F0 ðsÞ þ F1 ðsÞeT1 s þ F2 ðsÞeT2 s þ . . . (2.4.45)

where F0 ðsÞ; F1 ðsÞ; F2 ðsÞ; . . . are the Laplace transforms of f0 ðtÞ; f1 ðtÞ; f2 ðtÞ; . . .,
respectively.

Example 2.4.14 Deriving the Laplace transform of a rectangular pulse

Consider the rectangular pulse f ðtÞ in Figure 2.4.2(a), which mathematically is given by

F0 ; 0 ≤ t ≤ T
f ðtÞ ¼
0; t > T

It is easy to show that f ðtÞ can be expressed in a single expression

f ðtÞ ¼ F0 uðtÞ  F0 uðt  TÞ

where uðtÞ is the unit step function defined in Eq. (2.4.5), uðt  TÞ is the time-shifted unit step
function given in Example 2.4.4, and T is the time-delay parameter. Graphically, f ðtÞ is illustrated
in Figure 2.4.3(a). The Laplace transform of the rectangular pulse is

F0  
FðsÞ ¼ LfF0 uðtÞ  F0 uðt  TÞg ¼ 1  eTs
s

where Eq. (2.4.15) has been used.

62 2 Fundamentals of Mathematics

f(t) f(t)

F0 F0

f(t) =

0 0
t t
T
(a)

q(t ) q(t ) q(t )

q0
α α
q(t) =
1 1
0 0 0
t t t
T T

(b)

Figure 2.4.3 Graphical derivation of piecewise-continuous functions described in Figure 2.4.2: (a) rect-
angular pulse, (b) triangular pulse

Example 2.4.15 Deriving the Laplace transform of a triangular pulse

Consider the triangular pulse qðtÞ in Figure 2.4.2(b), which is defined by

αt; 0 ≤ t ≤ T
qðtÞ ¼
0; t > T

with the slope α ¼ q0 =T. The pulse can be written as

qðtÞ ¼ α t uðtÞ  αðt  TÞ uðt  TÞ  q0 uðt  TÞ

which is a linear combination of a ramp function (α t), a delayed ramp function (α ðt  TÞ), and a
delayed step function (q0 uðt  TÞ), as shown in Figure 2.4.3(b). The Laplace transform of this
triangular pulse is
α  q0
QðsÞ ¼ 2
1  eTs  eTs :
s s
 2.5 Differential Equations 63

2.5 Differential Equations

Differential equations are widely used to describe the behaviors of physical systems that arise
in various applications, such as heat and mass transfer, wave propagation, mechanical
vibrations, elasticity, combustion, fluid mechanics, gas dynamics, chemical reactions, elec-
trical circuitry, micro-electro-mechanical systems, satellite navigation, robotic systems, and
feedback controls. The derivation and solution of differential equations is especially import-
ant in modeling and analysis of dynamic systems.
A differential equation is an equation that contains an unknown function and its deriva-
tives. Differential equations can be categorized in several ways, which is briefly discussed in
the following subsections.

2.5.1 Linear and Nonlinear Differential Equations

Let uð xÞ be a function of x, where x is an independent variable. A linear differential equation
about uð xÞ is one which only involves a linear combination of uð xÞ and its derivatives, such as

d2 u du
2
þ 2 þ 4u ¼ 0
dx dx
du
þ ð1 þ 0:3ex Þu ¼ 5 cos x
dx

where the coefficients in linear combination are known functions of x. A nonlinear differential
equation of uð xÞ is an equation that contains products and/or nonlinear functions of uð xÞ and
its derivatives, such as

d2 u du
2
þ 2u þ 4u ¼ 0
dx
 dx
2
d2 u 2 du xu
þx þ pffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ 5 cos x
dx 2 dx 1 þ u2

In modeling dynamic systems, a general form of linear differential equations is

dn yðtÞ dn1 yðtÞ dyðtÞ

an ðtÞ n
þ an1 ðtÞ þ . . . þ a1 ðtÞ þ a0 ðtÞyðtÞ
dt dtn1 dt (2.5.1)
dn rðtÞ dm1 rðtÞ drðtÞ
¼ bm ðtÞ m þ bm1 ðtÞ þ . . . þ b1 ðtÞ þ b0 ðtÞrðtÞ
dt dtm1 dt

where yðtÞ is an unknown function, rðtÞ is a given function, t is an independent temporal

variable, and ak and bi are known functions of t. Equation (2.5.1) is called an n-th-order
differential equation because the highest-order derivative of y in it is of order n. The yðtÞ and
rðtÞ are also called the output and input of the dynamic system. If rðtÞ ¼ 0, the differential
equation is called homogeneous, and the corresponding solution yðtÞ is called a free response
64 2 Fundamentals of Mathematics

of the dynamic system. If rðtÞ ≠ 0, the differential equation is nonhomogeneous, and, provided
all initial conditions are zero, the solution yðtÞ of Eq. (2.5.1) is a forced response of the
dynamic system. If the ak values are all constants, the dynamic system is called a time-
invariant system. If any of the ak values are a function of t, the dynamic system is called a
time-varying system.
For convenience of analysis, Eq. (2.5.1) is cast as
 
an ðtÞDn þ an1 ðtÞDn1 þ . . . þ a1 ðtÞD þ a0 ðtÞ yðtÞ
  (2.5.2)
¼ bm ðtÞDm þ bm1 ðtÞDm1 þ . . . þ b1 ðtÞD þ b0 ðtÞ rðtÞ

where D is the differential operator given by

d
D¼ (2.5.3)
dt

Equation (2.5.2) can be further written in a compact form

AðDÞyðtÞ ¼ BðDÞrðtÞ (2.5.4)

where AðDÞ and BðDÞ are polynomials of the operator D, and they are given by

X
n
AðDÞ ¼ ak ðtÞDk
k¼0

X
m
BðDÞ ¼ bi ðtÞDi (2.5.5)
i¼0

Example 2.5.1 Modeling a spring–mass–damper dynamic system

Application of Newton’s second law of motion to the spring–mass–damper system in Figure 2.5.1
results in the second-order linear differential equation, which constitutes a model of the given
dynamic system.
This second-order linear differential equation is

_ þ kxðtÞ ¼ f ðtÞ
x ðtÞ þ cxðtÞ
m€

where xðtÞ is an unknown displacement, f ðtÞ is a given external force, t is a temporal variable, m, c,
and k are the constants that describe mass, damping coefficient, and spring coefficient, respect-
ively, and the over-dot stands for differentiation with respect to t, x_ ¼ dx=dt. The derivation of the
above equation will be presented in Chapter 3. This differential equation is also shown in Example
2.4.7. The differential equation can be written as

AðDÞxðtÞ ¼ BðDÞf ðtÞ

 2.5 Differential Equations 65

with
d
AðDÞ ¼ mD2 þ cD þ k; BðDÞ ¼ 1; with D ¼
dt

The spring–mass–damper system is a time-invariant system.

Figure 2.5.1 Spring–mass–damper dynamic system

2.5.2 Systems of Differential Equations

The differential equations in the previous section are single differential equations with one
unknown function. A system of differential equations involves multiple unknown functions
that are presented in coupled equations.

Example 2.5.2 Modeling of the two-mass dynamic system

Consider two masses (m1 and m2 ) that are coupled by a spring of coefficient k (see Figure 2.5.2).
Applying Newton’s second law of motion to this system, we obtain the governing differential
equations:
 
m1 x€1 ðtÞ þ k x1 ðtÞ  x2 ðtÞ ¼ f1 ðtÞ
 
m2 x€2 ðtÞ þ k x2 ðtÞ  x1 ðtÞ ¼ f2 ðtÞ

where x1 ðtÞ and x2 ðtÞ are the displacements of the masses, and f1 ðtÞ and f2 ðtÞ are external forces
applied to the masses. The derivation of these differential equations will be presented in Chapter 3.
By comparison, the single differential equation in Example 2.5.1 has one unknown function xðtÞ
while the coupled differential equations for the two-mass system have two unknown functions,
x1 ðtÞ and x2 ðtÞ.
In many engineering applications, a system of linear differential equations with n unknown
functions x1 ðtÞ; x2 ðtÞ; . . . ; xn ðtÞ can be written in the following matrix form

Am ðtÞDm þ Am1 ðtÞDm1 þ . . . þ A1 ðtÞD þ A0 ðtÞgxðtÞ ¼ f ðtÞ (2.5.6)

where D ¼ d=dt; m is the highest order of differentiation; xðtÞ and f ðtÞ are the vector of unknown
functions and the vector of given forcing functions, respectively, that are given by
66 2 Fundamentals of Mathematics

0 1 0 1
x1 ðtÞ f1 ðtÞ
Bx2 ðtÞ C Bf2 ðtÞ C
B C B C
xðtÞ ¼ B . C; f ðtÞ ¼ B . C (2.5.7)
@ .. A @ .. A
xn ðtÞ fn ðtÞ

and A0 ðtÞ; A1 ðtÞ; . . . ; Am ðtÞ are n  n matrices of known functions. For a set of linear differential
equations with constant coefficients, its matrix form is

Am Dm þ Am1 Dm1 þ . . . þ A1 D þ A0 gxðtÞ ¼ f ðtÞ (2.5.8)

in which all Ak are constant matrices. For instance, the differential equations in Example 2.5.2 are
of the matrix form
  
m1 0 x€1 ðtÞ k k x1 ðtÞ f ðtÞ
þ ¼ 1
0 m2 x€2 ðtÞ k k x2 ðtÞ f2 ðtÞ

or
M x€ðtÞ þ KxðtÞ ¼ f ðtÞ

with
 
x1 ðtÞ f1 ðtÞ m1 0 k k
xðtÞ ¼ ; f ðtÞ ¼ ; M¼ ; K¼ :
x2 ðtÞ f2 ðtÞ 0 m2 k k

Figure 2.5.2 Two-mass dynamic system

2.5.3 Partial Differential Equations

The differential equations in the previous subsections are called ordinary differential equa-
tions because the unknown functions in them are functions of one independent variable (like
x or t). In engineering problems there are also differential equations with unknowns being
functions of two or more independent variables. Two examples are given below.

(i) Equation of motion of an Euler–Bernoulli beam in transverse vibration

 2.5 Differential Equations 67


∂2 wðx; tÞ ∂2 ∂2 wðx; tÞ
ρðxÞ þ EIðxÞ ¼ f ðx; tÞ; 0 < x < L
∂t2 ∂x2 ∂x2

where wðx; tÞ, ρðxÞ, EIðxÞ, and L represent the transverse displacement, linear density (mass
per unit length), bending stiffness, and length of the beam, respectively; f ðx; tÞ is an external
force; and x and t are spatial and temporal parameters.

(ii) Governing equation for heat conduction in a three-dimensional body (heat equation)
 2
∂T ∂ T ∂2 T ∂2 T
cp ρ ¼κ þ þ þ qðx; y; z; tÞ
∂t ∂x2 ∂y2 ∂z2

where T ¼ Tðx; y; z; tÞ is the temperature distribution; q ¼ qðx; y; z; tÞ is the internal energy

generation; ρ; cp ; and κ are the volume density, specific heat, and thermal conductivity of the
body, respectively; x, y, and z are spatial coordinates; and t is a temporal parameter.
The scope of this textbook is about dynamic systems described by ordinary differential
equations. Consequently, derivation and solution of partial differential equations will not be
covered.

2.5.4 State Equations

State equations are often used in modeling, simulation, and analysis of dynamic systems with
multiple inputs and multiple outputs. State equations are a set of coupled first-order ordin-
ary differential equations of the form
8
>
> x_ 1 ¼ f1 ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; up Þ
>
<x_ 2 ¼ f2 ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; up Þ
.. (2.5.9)
>
> .
>
:
x_ n ¼ fn ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; up Þ

where the unknown functions xk ¼ xk ðtÞ; k ¼ 1; 2; . . . ; n are state variables; ui ¼ ui ðtÞ;

i ¼ 1; 2; . . . p are given inputs; and f1 ; f2 ; > . . . ; fn are functions of the state variables
and inputs. Equation (2.5.9) is a nonlinear form of a set of state equations. Such a set of
linear state equations is of the form
8
>
> x_ 1 ¼ a11 x1 þ a12 x2 þ . . . þ a1n xn þ b11 u1 þ b12 u2 þ . . . þ b1p up
>
<x_ 2 ¼ a21 x1 þ a22 x2 þ . . . þ a2n xn þ b21 u1 þ b22 u2 þ . . . þ b2p up
.. (2.5.10)
>
> .
>
:
x_ n ¼ an1 x1 þ an2 x2 þ . . . þ ann xn þ bn1 u1 þ bn2 u2 þ . . . þ bnp up

where ajk and bji are either known functions of t or constants.

68 2 Fundamentals of Mathematics

It should be noted that, in a state equation, the left-hand side only contains the first
derivative of a state variable and the right-hand side does not have any derivative term of xk
or ui .

Example 2.5.3 Deriving state equations

8
<x_ 1 ¼ 2ex1 cosð3x2 Þ þ 4u
2ex2 sinð3x2 Þ
: x_ 2 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi2
1þu

is a set of nonlinear state equations, and

8
<x_ 1 ¼ 2x1 þ 5x2  7x2 þ 6u1 þ 3u1  3u2
x_ ¼ 4x1 þ ð1 þ 0:2 sin 3tÞx2 þ 3x3 þ u2
: 2
x_ 3 ¼ 6x1 þ 10e0:25t x3 þ 13u1  9u2

is a set of linear state equations.

State equations are often cast in a matrix form. The nonlinear state equations (2.5.9) can be
written as

_ ¼ φðxðtÞ; uðtÞÞ
xðtÞ (2.5.11)

where xðtÞ and ϕ are given by

0 1 0 1 0 1
x1 ðtÞ u1 ðtÞ φ1 ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; up Þ
Bx2 ðtÞ C Bu2 ðtÞ C Bφ2 ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; up Þ C
B C B C B C
xðtÞ ¼ B . C; uðtÞ ¼ B .. C; φðxðtÞ; uðtÞÞ ¼ B .. C (2.5.12)
@ .. A @ . A @ . A
xn ðtÞ up ðtÞ φn ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; up Þ

with φ1 ; φ2 ; . . . φn being given nonlinear functions. The xðtÞ is called the state vector, and uðtÞ the
input vector. Likewise, the linear state equations (2.5.10) can be written as

_ ¼ AxðtÞ þ BuðtÞ
xðtÞ (2.5.13)

where xðtÞ and uðtÞ are the state and input vectors as given in Eq. (2.5.12), and matrices A and B are
as follows:
2 3 2 3
a11 a12 ... a1n b11 b12 ... b1p
6a21 a22 ... a2n 7 6b21 b22 ... b2p 7
6 7 6 7
A ¼ 6 .. .. .. 7; B ¼ 6 .. .. .. 7 (2.5.14)
4 . . ⋱ . 5 4 . . ⋱ . 5
an1 an2 ... ann bn1 bn2 ... bnp

The linear state equation (2.5.13) is a special case of the nonlinear equation (2.5.11) with
φ ¼ AxðtÞ þ BuðtÞ.
 2.5 Differential Equations 69

An n-th-order differential equation can be converted into a set of n state equations through the
proper selection of state variables. For instance, consider the governing equation of the spring–
mass–damper system in Example 2.5.1:

_ þ kxðtÞ ¼ f ðtÞ
x ðtÞ þ cxðtÞ
m€

_ the second-order differential equation

Through selection of state variables as x1 ¼ x and x2 ¼ x,
is cast into a system of state equations
(
x_ 1 ¼ x2
k c 1
x_ 2 ¼  x1  x2 þ f ðtÞ
m m m

The deduction of state equations from nonlinear differential equations can be similarly done. For
example, let the governing equation for a spring–mass–system subject to nonlinear damping be

_ þ c1 x_ 3 ðtÞ þ kxðtÞ ¼ f ðtÞ

x ðtÞ þ cxðtÞ
m€

where the cubic term c1 x_ 3 describes the damping nonlinearity. With the state variables defined by
_ the nonlinear state equations are obtained as follows:
x1 ¼ x and x2 ¼ x,
(
x_ 1 ¼ x2
k c c1 1
x_ 2 ¼  x1  x2  x32 þ f ðtÞ
m m m m

In general, to convert an n-th-order differential equation into a set of state equations, n state
variables must be defined. For the n-th-order linear differential equation
dn yðtÞ dn1 yðtÞ dyðtÞ
an ðtÞ n
þ an1 ðtÞ n1
þ . . . þ a1 ðtÞ þ a0 ðtÞyðtÞ ¼ rðtÞ (2.5.15)
dt dt dt

where ak ðtÞ are known functions, the selection of state variables by

x1 ðtÞ ¼ yðtÞ; x2 ðtÞ ¼ x_ 1 ðtÞ; x3 ðtÞ ¼ x_ 3 ðtÞ; . . . ; xn ðtÞ ¼ x_ n1 ðtÞ (2.5.16)

eventually yields the matrix state equation

_ ¼ AðtÞxðtÞ þ BrðtÞ
xðtÞ (2.5.17)

with matrices A and B being

2 3 0 1
0 1 0 ... 0 0
6 0 0 1 ... 0 7 B C
0
6 .. .. .. .. 7 B C
..
6 7 B C
AðtÞ ¼ 6 . . . ⋱ . 7; B ¼ B C. (2.5.18)
6 7 B C
6 0 0 0 ... 1 7 B 0 C
4 a0 ðtÞ a1 ðtÞ a2 ðtÞ an1 ðtÞ 5 @ 1 A
   ... 
an ðtÞ an ðtÞ an ðtÞ an ðtÞ an ðtÞ

The derivation of state equations for dynamic systems will be further discussed later on.
70 2 Fundamentals of Mathematics

2.6 Solution of Linear Ordinary Differential Equations with Constant

Coefficients
2.6.1 Free Response and Forced Response
In this section, we present analytical methods for the solution of linear ordinary differential
equations with constant coefficients, which, as mentioned previously, are used to model
time-invariant dynamic systems. The solution of linear differential equations with noncon-
stant coefficients and nonlinear equations will be discussed in numerical integration of state
equations (see Section 2.8).
The application of the methods described in this section to different types of dynamic
systems is further discussed in Chapters 3 to 9.
The differential equation under consideration has the following form

AðDÞyðtÞ ¼ BðDÞrðtÞ (2.6.1)

and is subject to the initial conditions

yð0Þ ¼ a0;0 ; Dyð0Þ ¼ a0;1 ; D2 yð0Þ ¼ a0;2 ; . . . ; Dn1 yð0Þ ¼ a0;n1 (2.6.2)

where D ¼ d=dt, and AðDÞ and BðDÞ are polynomials of the operator D, such as

X
n X
m
AðDÞ ¼ ak Dk ; BðDÞ ¼ bi Di (2.6.3)
k¼0 i¼0

where ak and bi are constant coefficients, and a0; 0 ; a0; 1 ; . . . ; a0; n1 are the prescribed initial
values. The given function rðtÞ is normally referred to as an input or a forcing function. In
general, for a physically viable system, n≥m. Thus, the mathematical problem herein is to
solve the differential equation (2.6.1) for the unknown function yðtÞ, for given input rðtÞ and
initial disturbances described in Eq. (2.6.2).
According to the theory of differential equations, the solution of the above-described
equation with the initial conditions can be written as

yðtÞ ¼ yI ðtÞ þ yF ðtÞ (2.6.4)

where yI ðtÞ is the solution of the homogeneous equation

AðDÞyðtÞ ¼ 0 (2.6.5)

subject to the initial conditions of Eq. (2.6.2), and yF ðtÞ is the solution of Eq. (2.6.1) subject to
the homogeneous initial conditions, such as

yð0Þ ¼ 0; Dyð0Þ ¼ 0; D2 yð0Þ ¼ 0; . . . ; Dn1 yð0Þ ¼ 0 (2.6.6)

 2.6 Solution of Linear Ordinary Differential Equations 71

In modeling and analysis of dynamic systems, yI ðtÞ and yF ðtÞ are called the free response
and forced response respectively. Physically, yI ðtÞ is the system response caused by the initial
disturbances prescribed by Eq. (2.6.2), and yF ðtÞ is caused by the input rðtÞ. Hence, the
complete solution of a differential equation can be viewed as a sum of free response and
forced response.

Example 2.6.1 Response of the spring–mass–damper system

Consider the spring–mass–damper system given in Example 2.5.1.
The system response, namely the displacement xðtÞ of the mass, is governed by the differential
equation

_ þ kxðtÞ ¼ f ðtÞ
x ðtÞ þ cxðtÞ
m€

subject to the initial conditions

_
xð0Þ ¼ x0 ; xð0Þ ¼ v0

where x0 and v0 are the initial displacement and the initial velocity of the mass, respectively. The
solution of this governing differential equation can be written as

xðtÞ ¼ xI ðtÞ þ xF ðtÞ

in which the free response xI ðtÞ is the solution of

_ þ kxðtÞ ¼ 0
x ðtÞ þ cxðtÞ
m€
_
xð0Þ ¼ x0 ; xð0Þ ¼ v0

and the forced response xF ðtÞ is the solution of

_ þ kxðtÞ ¼ f ðtÞ
x ðtÞ þ cxðtÞ
m€
_
xð0Þ ¼ 0; xð0Þ ¼0

There exist four widely used analytical methods for the solution of linear differential
equations with constant coefficients: (1) direct integration, (2) the method of separation of
variables, (3) the Laplace transform, and (4) the method of undetermined coefficients.

2.6.2 Solution by Direct Integration

For a linear differential equation with constant coefficients, where it is possible to isolate the
function derivative on the left-hand side

Dn yðtÞ ¼ f ðtÞ (2.6.7)

the solution can be found by integrating both sides of the equation n times
72 2 Fundamentals of Mathematics

ðt  ðt  ðt
yðtÞ ¼ yð0Þ þ _
yð0Þ þ ... yðn1Þ ð0Þ þ f ðtÞ dt dt . . . dt (2.6.8)
0 0 0

This method is straightforward, but the inherent complexity of the integration operation
limits its applicability to simple first- and second-order differential equations.

Example 2.6.2 Solving a first-order differential equation by direct integration

Consider a dynamic system governed by the first-order differential equation with the given initial
conditions

_ ¼ f ðtÞ
mxðtÞ
xð0Þ ¼ x0

Integrate both sides of this equation once to derive the solution

ð
xðtÞ ðt
dx 1
dt ¼ f ðtÞdt
dt m
xð0Þ 0

ðt
1
xðtÞ ¼ x0 þ f ðtÞdt
m
0

If a dynamic system is governed by the second-order differential equation with the given initial
conditions
m€x ðtÞ ¼ f ðtÞ
_
xð0Þ ¼ x0 ; xð0Þ ¼ v0

integration needs to be done twice as follows:

_ð
xðtÞ ðt
d2 x 1
dt ¼ f ðtÞdt
dt2 m
_
xð0Þ 0

ðt
1
_ ¼ v0 þ
xðtÞ f ðtÞdt
m
0
0 1
ð
xðtÞ ðt ðt
dx 1
dt ¼ @v0 þ f ðtÞdtA dt
dt m
xð0Þ 0 0
0 1
ðt ðt
1 @
xðtÞ ¼ x0 þ v0 t þ f ðtÞdtA dt
m
0 0

The following numerical example illustrates the direct integration method.

 2.6 Solution of Linear Ordinary Differential Equations 73

Example 2.6.3 Solving a second-order differential equation by direct integration

Consider a dynamic system governed by the second-order differential equation with the given
initial conditions

x€ðtÞ ¼ 3t2
_
xð0Þ ¼ 1; xð0Þ ¼2

The solution is found by direct integration as follows:

_ð
xðtÞ ðt
d2 x
dt ¼ 3t2 dt
dt2
_
xð0Þ 0

_ ¼ 2 þ t3
xðtÞ
ð
xðtÞ ðt
dx  
dt ¼ 2 þ t3 dt
dt
xð0Þ 0

1
xðtÞ ¼ 1 þ 2t þ t4
4

2.6.3 Solution by the Separation of Variables

This approach can be viewed as an extension of the direct integration method, mostly
applicable to first-order differential equations of the following structure:

dx
¼ f1 ðtÞf2 ðxÞ (2.6.9)
dt

For the first-order equation shown above, the solution is found in the following way. First,
the equation is reformulated in such a way that its left-hand side is only in terms of x, while
the right-hand side is only in terms of t:

dx
¼ f1 ðtÞdt (2.6.10)
f2 ðxÞ

Then, both sides of the resulting equation are integrated:

ð
xðtÞ ðt
dx
¼ f1 ðtÞdt (2.6.11)
f2 ðxÞ
xð0Þ 0
74 2 Fundamentals of Mathematics

The sought solution xðtÞ is found by Eq. (2.6.11), provided that integrals on the right-hand
side and left-hand side can be evaluated analytically, and xðtÞ can be derived from the
resulting expression.

Example 2.6.4 Solving a first-order differential equation by the separation

of variables
Consider a dynamic system governed by the first-order differential equation with the given initial
conditions

_ þx¼3
xðtÞ
xð0Þ ¼ 2

Rearrange this equation into the form, required by the method of separation of variables
dx
¼3x
dt

Obviously, f1 ðtÞ ¼ 1 and f2 ðxÞ ¼ 3  x. Integrating both sides, we obtain

ð
xðtÞ ðt
dx
¼ 1dt
3x
xð0Þ 0

Evaluation of the integral on the right-hand side of the equation is straightforward, while the
integral on the left-hand side can be evaluated using integral tables:
ð ð

xðtÞ xðtÞ
1



xðtÞ

ðt
dx dx




¼ ¼ ln
3  x

¼ ln
3  xðtÞ
¼ 1dt ¼ t
3x 3  x 1

xð0Þ 2 2 0

From the resulting expression xðtÞ can be easily derived:

lnj3  xðtÞj ¼ t
3  xðtÞ ¼ et
xðtÞ ¼ 3  et

2.6.4 Solution by the Laplace Transform

The Laplace transform is a widely used method in modeling, analysis, and design of control
algorithms for linear time-invariant dynamic systems, and in solving the related differential
equations with constant coefficients. As mentioned earlier in this chapter (see Section 2.4),
the Laplace transform reduces the original problem, described by differential equations, to a
 2.6 Solution of Linear Ordinary Differential Equations 75

set of algebraic equations in the s-domain. The functions of s, which result from solving these
algebraic equations, are then transformed into the time domain by using the inverse Laplace
transform. To understand how this approach works, consider a simple example:

_ þ 3xðtÞ ¼ 5
xðtÞ
xð0Þ ¼ 2

Taking the Laplace transform of the above equation gives

1 1
sX ðsÞ  xð0Þ þ 3X ðsÞ ¼ 5 ) ðs þ 3ÞX ðsÞ ¼ 5 þ 2
s s

where the initial condition xð0Þ ¼ 2 has been used. It follows that the s-domain response is

2s þ 5
X ðsÞ ¼
sðs þ 3Þ

By partial fraction expansion (see Section 2.4.4), the solution (time-domain response) to the
differential equation is
   
5 1 1 1 1  
xðtÞ ¼ L1 þ ¼ 5 þ e3t
3 s 3 sþ3 3

The above example implies three basic steps in the solution of general differential equa-
tions via the Laplace transformation, which are outlined below.

Step 1: Laplace Transform of a Differential Equation

Take the Laplace transform of Eq. (2.6.1) and substitute the initial conditions (2.6.2) into the
resulting equation. This leads to

AðsÞYðsÞ ¼ BðsÞRðsÞ þ IðsÞ (2.6.12)

where YðsÞ and RðsÞ are the Laplace transforms of yðtÞ and rðtÞ, respectively; AðsÞ and BðsÞ
are the polynomials in Eq. (2.6.3) with operator D replaced by the Laplace transform
parameter s:

X
n X
m
AðsÞ ¼ ak sk ; BðsÞ ¼ bi si (2.6.13)
k¼0 i¼0

and IðsÞ is an expression consisting of the initial values of yðtÞ; rðtÞ and its derivatives. In
general, IðsÞ is a polynomial of s and IðsÞ ¼ 0 when all the initial values of yðtÞ; rðtÞ, and their
derivatives are set to zero. For instance, consider the Laplace transform of the differential
equation
76 2 Fundamentals of Mathematics

_ þ kxðtÞ ¼ f ðtÞ
x ðtÞ þ cxðtÞ
m€
_
xð0Þ ¼ x0 ; xð0Þ ¼ v0

in Example 2.5.1, which yields Eq. (2.6.12) with

AðsÞ ¼ ms2 þ cs þ k; BðsÞ ¼ 1; IðsÞ ¼ mx0 s þ mv0 þ cx0

Note that IðsÞ ¼ 0 if and only if x0 ¼ 0 and v0 ¼ 0.

Step 2: Determination of the s-Domain Response

By Eq. (2.6.12), the s-domain solution is of the form

BðsÞRðsÞ þ IðsÞ
YðsÞ ¼ (2.6.14)
AðsÞ

Step 3: Determination of the Time-Domain Response via the Inverse Laplace Transform
The inverse Laplace transform of Eq. (2.6.14) gives

yðtÞ ¼ yI ðtÞ þ yF ðtÞ (2.6.15)

where
 
1 IðsÞ 1 BðsÞ
yI ðtÞ ¼ L ; yF ðtÞ ¼ L RðsÞ (2.6.16)
AðsÞ AðsÞ

with yI ðtÞ and yF ðtÞ being the free response and forced response as described in Section 2.6.1.
The inverse Laplace transformation in Eq. (2.6.16) can be performed via partial fraction
expansion as described in Section 2.4.4.

Example 2.6.5 Solving a second-order differential equation by the Laplace transform

Consider the differential equation

m€x ðtÞ þ kxðtÞ ¼ f0

_
xð0Þ ¼ x0 ; xð0Þ ¼ v0

which is a special case of the Example 2.6.1 with c ¼ 0 and f ðtÞ being a step (constant) input. Using
the results in Example 2.5.1, the s-domain solution is

mx0 s þ mv0 f0 x 0 s þ v0 f0 1 s
X ðsÞ ¼ þ ¼ þ  2
ms þ k
2 sðms þ kÞ s þ ωn k s s þ ω2n
2 2 2
 2.6 Solution of Linear Ordinary Differential Equations 77

k
where ω2n ¼ . With Table 2.4.1, the inverse Laplace transform of the above equation gives the
m
solution of the differential equation:

v0 f0
xðtÞ ¼ L1 fX ðsÞg ¼ x0 cos ωn t þ sin ωn t þ ð1  cos ωn tÞ
ωn k

According to Section 2.6.1, the free response and forced response of the differential equation are
v0
xI ðtÞ ¼ x0 cos ωn t þ sin ωn t
ωn
f0
xF ðtÞ ¼ ð1  cos ωn tÞ
k

Example 2.6.6 Solving a third-order differential equation by the Laplace transform

A third-order differential equation with initial conditions is given as follows:

€x˙ðtÞ þ 6€ _ þ 6xðtÞ ¼ 6ð1  e4t Þ

x ðtÞ þ 11xðtÞ
_
xð0Þ ¼ 1; xð0Þ ¼ 0; x€ð0Þ ¼ 0

The Laplace transform of the above equation leads to


  1 1 s4 þ 10s3 þ 34s2 þ 44s þ 4
s þ 6s þ 11s þ 6 X ðsÞ ¼ 6 
3 2
þ s2 þ 6s þ 11 ¼
s sþ4 sðs þ 4Þ

It can be shown that the algebraic equation s3 þ 6s2 þ 11s þ 6 ¼ 0 has roots 1; 2; and  3,
which indicates that s3 þ 6s2 þ 11s þ 6 ¼ ðs þ 1Þðs þ 2Þðs þ 3Þ. From this result it follows that

s4 þ 10s3 þ 34s2 þ 44s þ 24 s4 þ 10s3 þ 34s2 þ 44s þ 24

X ðsÞ ¼ ¼
s5 þ 10s þ 35s þ 50s þ 24s sðs þ 1Þðs þ 2Þðs þ 3Þðs þ 4Þ
4 3 2

By partial fraction expansion (Section 2.4.4),

1 5 2 3 1
X ðsÞ ¼  þ  þ
s 6ðs þ 1Þ s þ 2 2ðs þ 3Þ 3ðs þ 4Þ

Thus, the sought solution (time-domain response) of the given differential equation is
5 3 1
xðtÞ ¼ L1 fX ðsÞg ¼ 1  et þ 2e2t  e3t þ e4t
6 2 3

The xðtÞ is plotted against t in Figure 2.6.1.

78 2 Fundamentals of Mathematics

1.05

1.

x(t )
0.95

0.9

0.85
0 1 2 3 4 5 6 7 8 9 10
Time (seconds)

Figure 2.6.1 Time-domain response of the given third-order differential equation

Example 2.6.7 Solving a first-order differential equation with a piecewise-

continuous forcing function by the Laplace transform
The Laplace transform is quite useful for solving linear time-invariant differential equations with
a piecewise-continuous forcing function.
Consider the system with the rectangular pulse forcing function (see Figure 2.4.2(a)) with the
amplitude F0 ¼ 3 and duration T ¼ 2:

_ þ 4xðtÞ ¼ f ðtÞ ¼ 3uðtÞ  3uðt  2Þ for t > 0; subject to xð0Þ ¼ 1

xðtÞ

Taking the Laplace transform of this equation, we obtain (the Laplace transform of the rectangu-
lar pulse was derived in Example 2.4.14):

_ þ 4xðtÞg ¼ sX ðsÞ  xð0Þ þ 4X ðsÞ ¼ ðs þ 4ÞX ðsÞ  1

LfxðtÞ
3 
¼ Lf3uðtÞ  3uðt  2Þg ¼ 1  e2s
s

Then the solution in the Laplace domain is derived from this algebraic equation as

3  3e2s þ s 3 3 1
X ðsÞ ¼ ¼  e2s þ
sðs þ 4Þ sðs þ 4Þ sðs þ 4Þ sþ4

To invert this solution to the time domain, use the partial fraction expansion, Table 2.4.1, and
the Laplace transform properties
 2.6 Solution of Linear Ordinary Differential Equations 79


3 3 1
xðtÞ ¼ L1 fX ðsÞg ¼ L 1  e2s þ
sðs þ 4Þ sðs þ 4Þ sþ4
   
3 3 1 3 1 3 3 3 
L 1 ¼ L1  ¼ uðtÞ  e4t uðtÞ ¼ 1  e4t uðtÞ
sðs þ 4Þ 4 s 4 sþ4 4 4 4

3 3 
L1 e2s ¼ 1  e4ðt2Þ uðt  2Þ
sðs þ 4Þ 4

1
L1 ¼ e4t uðtÞ
sþ4

3  3 
xðtÞ ¼ e4t þ 1  e4t uðtÞ  1  e4ðt2Þ uðt  2Þ
4 4

which can also be written as

8
> 3 
< e4t þ 1  e4t for 0 ≤ t ≤ 2
xðtÞ ¼ 4  
>  
: e4t þ 3 1  e4t þ 3 1  e4ðt2Þ for t > 2
4 4

This piecewise-continuous solution is shown in Figure 2.6.2.

1.0

0.8

0.6
x (t )

0.4

0.2

0.0

0 1 2 3 4
Time (seconds)

Figure 2.6.2 Piecewise-continuous time-domain response of the given first-order differential

equation
80 2 Fundamentals of Mathematics

2.6.5 The Method of Undetermined Coefficients

Consider the n-th-order differential equation

an Dn þ an1 Dn1 þ . . . þ a1 D þ a0 gxðtÞ ¼ f ðtÞ (2.6.17)

subject to the initial conditions

xð0Þ ¼ a0;0 ; Dxð0Þ ¼ a0;1 ; . . . ; Dn1 xð0Þ ¼ a0;n1 (2.6.18)

with D ¼ d=dt. Here, for simplicity of presentation, f ðtÞ ¼ BðDÞrðtÞ has been used. In the
context of the method of undetermined coefficients, the solution of Eq. (2.6.17) is written as

xðtÞ ¼ xh ðtÞ þ xp ðtÞ (2.6.19)

where xh ðtÞ (homogenous solution) is a nonzero or nontrivial solution of the complementary

homogenous equation

an Dn þ an1 Dn1 þ . . . þ a1 D þ a0 gxh ðtÞ ¼ 0 (2.6.20)

and xp ðtÞ is a particular solution of Eq. (2.6.17). Here, neither xh ðtÞ nor xp ðtÞ have to satisfy the
initial conditions of Eq. (2.6.18).
It should be pointed out that the particular solution xp ðtÞ is not unique. For instance,
consider the first-order differential equation

_ þ 2xðtÞ ¼ 4
xðtÞ

It is easy to show that both xp1 ðtÞ ¼ 2 and xp2 ðtÞ ¼ 2 þ α e2t , with α being an arbitrary
number, satisfy the differential equation. Nevertheless, any particular solution that can be
used in the format of Eq. (2.6.19) will yield the unique solution of the differential equation
(2.6.17) subject to the initial conditions of Eq. (2.6.18).
It needs to be noted that the homogeneous solution xh ðtÞ and particular solution xp ðtÞ must
be linearly independent, meaning that the expression

α1 xh ðtÞ þ α2 xp ðtÞ ¼ 0

holds true if and only if α1 ¼ α2 ¼ 0

Solution of the Complementary Homogenous Equation

To determine the solution of the homogenous equation, write it in an exponential form

xh ðtÞ ¼ Aeλt (2.6.21)

and substitute it into Eq. (2.6.20). This leads to

 2.6 Solution of Linear Ordinary Differential Equations 81

an λn þ an1 λn1 þ . . . þ a1 λ þ a0 gAeλt ¼ 0

which, by the nontrivial solution assumption (A ≠ 0), yields the characteristic equation of the
differential equation:

an λn þ an1 λn1 þ . . . þ a1 λ þ a0 ¼ 0 (2.6.22)

Equation (2.6.17), being an n-th-order polynomial of λ, has n roots. Denote the character-
istic roots by λ1 ; λ2 ; . . . ; λn . The expression for xh ðtÞ must contain n linearly independent
terms. If all the characteristic roots are distinct, the solution of the homogenous equation is

xh ðtÞ ¼ A1 eλ1 t þ A2 eλ2 t þ . . . þ An eλn t (2.6.23)

which has n linearly independent terms. If there exist repeated roots, the expression for xh ðtÞ
will be different in order to have n linearly independent terms. For instance, assume that the
characteristic equation (2.6.22) has p identical roots of λ0 and m distinct roots of
λ1 ; λ2 ; . . . ; λm , where λ0 ≠ λk for k ¼ 1; 2; . . . ; m and p þ m ¼ n. Here, the integer p is called
root multiplicity. In this case,
 
xh ðtÞ ¼ A1 eλ1 t þ A2 eλ2 t þ . . . þ Am eλm t þ B0 þ B1 t þ . . . þ Bp1 tp1 eλ0 t (2.6.24)

where the last p terms on the right-hand side of the equation are related to λ0 of multiplicity p.
The coefficients A and B in Eqs. (2.6.23) and (2.6.24) will be determined later on through the
use of the initial conditions of Eq. (2.6.18).
The characteristic equation (2.6.22) can be written as

AðλÞ ¼ 0 (2.6.25)

where AðsÞ is the polynomial given in Eq. (2.6.13), with the s replaced by λ. This means that
the roots of the characteristic equation (2.6.22) are the same as the poles of the s-domain
solution depicted in Eq. (2.6.14).

Example 2.6.8 Deriving the general homogeneous solution of a linear differential

equation with constant coefficients and real distinct roots
Consider the differential equation in the Example 2.6.6, which has the complementary homoge-
neous equation

€x˙ðtÞ þ 6€ _ þ 6xðtÞ ¼ 0
x ðtÞ þ 11xðtÞ

and the characteristic equation

λ3 þ 6λ2 þ 11λ þ 6 ¼ 0

with roots λ1 ¼ 1; λ2 ¼ 2, and λ3 ¼ 3 The solution of the homogeneous equation then is

xh ðtÞ ¼ A1 et þ A2 e2t þ A3 e3t :

82 2 Fundamentals of Mathematics

Example 2.6.9 Deriving the general homogeneous solution of linear differential

equation with constant coefficients and real repeated roots
Consider the complementary homogeneous equation

€x˙ðtÞ þ 5€ _ þ 3xðtÞ ¼ 0
x ðtÞ þ 7xðtÞ

which has the characteristic equation

λ3 þ 5λ2 þ 7λ þ 3 ¼ 0

The characteristic roots are 1;  1; and  3, with root 1 having multiplicity 2. By Eq.
(2.6.24), the solution of the homogeneous equation is

xh ðtÞ ¼ Ae3t þ ðB0 þ B1 tÞet

Determination of a Particular Solution

In the context of the method of undetermined coefficients, an educated guess of the
form of a particular solution xp ðtÞ is made for the given forcing function f ðtÞ on the right-
hand side of Eq. (2.6.17). The guessed xp ðtÞ contains some coefficients that are to be solved
for.
In general, the guessed form of xp ðtÞ should “mirror” the form of the forcing function as
shown in Table 2.6.1 for first-order differential equations.
For the first-order differential equation with an exponential forcing function, the existence
of two options for a particular solution is prompted by the necessity of having linearly
independent xh ðtÞ and xp ðtÞ as explained in detail below.

First-Order Linear Ordinary Differential Equations

A first-order differential equation

_ þ bxðtÞ ¼ f ðtÞ; a ≠ 0
axðtÞ (2.6.26)

has the corresponding characteristic equation aλ þ b ¼ 0, which has a real root

λ1 ¼ b=a. Table 2.6.1 lists polynomial, sinusoidal, and exponential forms of forcing
function, and corresponding “guessed” particular solutions. In the table, the coeffi-
cients of f ðtÞ are given; the coefficients of xp ðtÞ are determined by plugging it into Eq.
(2.6.26). When f ðtÞ is an exponential function with α ¼ b=a (the characteristic root), the
particular solution is xp ðtÞ ¼ Bteαt . This is because the eat itself is a solution of the homoge-
_ þ bxðtÞ ¼ 0, which does not satisfy the inhomogeneous equation
neous equation axðtÞ
(2.6.26).
 2.6 Solution of Linear Ordinary Differential Equations 83

Example 2.6.10 Deriving a particular solution of a first-order linear differential

equation with constant coefficients
Consider the differential equation

_ þ 3xðtÞ ¼ 2ð1  e3t Þ

xðtÞ

the characteristic equation of which has the root 3. Note that the characteristic root satisfies the
condition α ¼ b=a. According to Table 2.6.1 and the principle of superposition, a particular
solution of the differential equation is

xp ðtÞ ¼ d0 þ Ate3t

Substituting the guessed form into the differential equation gives

   
Ae3t  3Ate3t þ 3 d0 þ Ate3t ¼ 2 1  e3t

which results in
3d0 þ Ae3t ¼ 2  e3t

Comparison of the coefficients of the relevant terms on the two sides of the equation yields
Constant term : 3d0 ¼ 2
Exponential term : A ¼ 2

Thus, with d0 ¼ 2=3 and A ¼ 2, the particular solution is

2
xp ðtÞ ¼  2te3t
3

Table 2.6.1 Particular solutions for first-order differential equations

Forcing function f ðtÞ Particular solution xp ðtÞ

(a) Polynomial xp ðtÞ ¼ dn tn þ dn1 tn1 þ . . . þ d1 t þ d0

f ðtÞ ¼ cn tn þ cn1 tn1 þ . . . þ c1 t þ c0
(b) Sinusoidal xp ðtÞ ¼ B1 sin ωt þ B2 sin ωt
f ðtÞ ¼ A1 sin ωt þ A2 cos ωt
(c) Exponential Case 1. If α ≠  b=a, xp ðtÞ ¼ Beat
f ðtÞ ¼ Aeat Case 2. If α ¼ b=a, xp ðtÞ ¼ Bteat
(coefficients a and b are the coefficients of the first-order
_ þ bxðtÞ ¼ f ðtÞ; a ≠ 0)
differential equation axðtÞ
84 2 Fundamentals of Mathematics

Given a general forcing function

f ðtÞ ¼ eat ½Ak ðtÞcos ωt þ Bm ðtÞsin ωt (2.6.27)

with Ak ðtÞ and Bm ðtÞ being polynomials of orders k and m, respectively, a particular solution
of Eq. (2.6.26) can be written as

xp ðtÞ ¼ tμ eat ½Qn ðtÞcos ωt þ Rn ðtÞsin ωt (2.6.28)

where Qn ðtÞ and Rn ðtÞ are polynomials of order n ¼ maxðk; mÞ such that

Qn ðtÞ ¼ qn tn þ qn1 tn1 þ . . . þ q1 t þ q0

(2.6.29)
Rn ðtÞ ¼ rn tn þ rn1 tn1 þ . . . þ r1 t þ r0

and μ is an integer that is given by

0; if α ≠  b=a
μ¼ (2.6.30)
1; if α ¼ b=a

The coefficients of the polynomials in Eq. (2.6.29) are to be determined. As can be seen, the
particular solutions in Table 2.6.1 are special cases of the form given in Eq. (2.6.28).
In the context of the method of undetermined coefficients, the principle of superposition for
linear differential equations can be applied. For example, if a forcing function can be written
as a linear combination of the functions given in Table 2.6.1 or the form of Eq. (2.6.28), such as

f ðtÞ ¼ 2t þ 3 cosð4tÞ

a particular solution can be written as

xp ðtÞ ¼ d1 t þ d0 þ A sinð4tÞ þ B cosð4tÞ

where the forms in the first and second rows of Table 2.6.1 have been used.

Second-Order Linear Ordinary Differential Equations

The second-order differential equation is of the form

a€ _ þ cxðtÞ ¼ f ðtÞ; a ≠ 0
x ðtÞ þ bxðtÞ (2.6.31)

The corresponding characteristic equation, by Eq. (2.6.22), is

aλ2 þ bλ þ c ¼ 0 (2.6.32)

with its roots given by

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b b2  4ac
ðλ1 ; λ2 Þ ¼   (2.6.33)
2a 2a
 2.6 Solution of Linear Ordinary Differential Equations 85

Table 2.6.2 Particular solutions for second-order differential equations

Forcing function f ðtÞ Particular solution xp ðtÞ

(a) Polynomial f ðtÞ ¼ cn tn þ cn1 tn1 þ . . . þ c1 t þ c0 xp ðtÞ ¼ dn tn þ dn1 tn1 þ . . . þ d1 t þ d0

(b) Sinusoidal f ðtÞ ¼ A1 sin ωt þ A2 cos ωt Case 1. If b ≠ 0 or if b ¼ 0 and ω2 ≠ c=a,
xp ðtÞ ¼ B1 sin ωt þ B2 cos ωt
Case 2. If b ¼ 0 and ω2 ¼ c=a,
xp ðtÞ ¼ tðB1 sin ωt þ B2 cos ωtÞ
(c) Exponential f ðtÞ ¼ Aeαt Case 1. If α ≠ λ1 and α ≠ λ2 ,
xp ðtÞ ¼ Beat
Case 2. If λ1 ≠ λ2 and α is equal to one of the roots,
xp ðtÞ ¼ Bteat
Case 3. If α ¼ λ1 ¼ λ2 ,
xp ðtÞ ¼ Bt2 eαt

Table 2.6.2 lists polynomial, sinusoidal, and exponential forms of the forcing function,
and corresponding “guessed” particular solutions.
As can be seen from Table 2.6.2, the form of a particular solution again “mirrors” the
forcing function and is relevant to the coincidence between the characteristic roots ðλ1 ; λ2 Þ and
certain parameters of a given forcing function. For instance, when b ¼ 0, Eq. (2.6.31)
becomes

x ðtÞ þ cxðtÞ ¼ f ðtÞ; a ≠ 0

a€ (2.6.34)

with the characteristic roots

pffiffiffiffiffiffiffi pffiffiffiffiffiffiffi
ðλ1 ; λ2 Þ ¼ j c=a; j ¼ 1 (2.6.35)

If f ðtÞ is sinusoidal with ω2 ¼ c=a, the particular solution by Table 2.6.2 has the form

xp ðtÞ ¼ tðB1 sin ωt þ B2 cos ωtÞ (2.6.36)

indicating an unbounded oscillatory response. This solution is commonly seen in mechanical

systems, and is often referred to as resonant vibration. One typical example of this phenom-
enon is seen in the Example 2.6.5 with zero damping, for which the governing differential
equation is

x ðtÞ þ kxðtÞ ¼ f ðtÞ

m€

and the characteristic roots are

pffiffiffiffiffiffiffiffiffi
ðλ1 ; λ2 Þ ¼ jωn ; with ωn ¼ k=m
86 2 Fundamentals of Mathematics

where ωn is called the natural frequency of the spring–mass system. Equation (2.6.36) implies
that a resonant response occurs when the frequency ω of the sinusoidal forcing function
matches the natural frequency of the system.
Showing that the amplitude of the resonant vibration is unbounded could be done by
solving for the coefficients B1 and B2 , and then taking the limit of the solution with time
approaching infinity.
To simplify the derivations, assume zero initial conditions and let the forcing function be
sinusoidal. Then,

m€x ðtÞ þ kxðtÞ ¼ α sinðωtÞ (2.6.37)

pffiffiffiffiffiffiffiffiffi
where ω ¼ ωn ¼ k=m and α 2 ℝ.
Plugging the derived particular solution (Eq. (2.6.36)) into the governing differential
equation, we obtain
 
 þ B2 2 k  mω t cosðωtÞ  2B2 mω sinðωtÞ
2
2B1 mω cosðωtÞ
 B1 k  mω t sinðωtÞ ¼ α sinðωtÞ

which yields the following system of equations for deriving the coefficients B1 and B2 :
8
>
> 2B1 mω ¼ 0 (
<2B mω ¼ α B1 ¼ 0
 2  ) B2 ¼  α ¼  pαffiffiffiffiffiffi
>B2 k  mω t ¼ 0 8t > 0
>
2
: 2mω 2 km
B1 k  mω2 t ¼ 0 8t > 0
pffiffiffiffiffiffiffiffiffi
Since ω ¼ ωn ¼ k=m, k  mω2 ≡ 0 and the last two equalities are always satisfied.
Hence, the solution of Eq. (2.6.37) is
α pffiffiffiffiffiffiffiffiffi
xðtÞ ¼ xp ðtÞ ¼  pffiffiffiffiffiffi t cosðt k=mÞ (6.2.38)
2 km

The upper and lower limits of Eq. (6.2.38) are þ∞ and ∞, respectively, i.e. the amplitude
of the resonant vibration is unbounded.
Using similar reasoning, the solution for the system with f ðtÞ ¼ α cosðωtÞ is
derived as
α pffiffiffiffiffiffiffiffiffi
xðtÞ ¼ xp ðtÞ ¼ pffiffiffiffiffiffi t sinðt k=mÞ (6.2.39)
2 km

The upper and lower limits of Eq. (6.2.39) are also þ∞ and ∞, respectively, showing that
this resonant vibration also has an unbounded amplitude.
A numerical example of such system is shown in Example 2.6.11.
 2.6 Solution of Linear Ordinary Differential Equations 87

Example 2.6.11 Deriving a particular solution of a second-order linear differential

equation with constant coefficients and complex roots
Consider the differential equation

x€ðtÞ þ 9xðtÞ ¼ 3 sinð3tÞ

Its characteristic equation is

λ2 þ 9 ¼ 0
pffiffiffiffiffiffiffi
with roots being j3, j ¼ 1. According to Table 2.6.2 or Eq. (2.6.36), a particular solution is
 
xp ðtÞ ¼ t B1 sinð3tÞ þ B2 cosð3tÞ

By using the differentiation formula

d2 d2 u du dv d2 v
ðuvÞ ¼ v þ 2 þ u (2.6.40)
dt2 dt2 dt dt dt2

we obtain
   
x€p ðtÞ ¼ 6 B1 cosð3tÞ  6B2 sinð3tÞ  9t B1 sinð3tÞ  6B2 cosð3tÞ
¼ 6B1 cosð3tÞ  6B2 sinð3tÞ  9xp ðtÞ þ 9xp ðtÞ ¼ 3 sinð3tÞ

Plugging the above expression into the originally given differential equation yields

6B1 cosð3tÞ  6B2 sinð3tÞ  9xp ðtÞ þ 9xp ðtÞ ¼ 3 sinð3tÞ

which is reduced to
6B1 cosð3tÞ  6B2 sinð3tÞ ¼ 3 sinð3tÞ

Comparison of relevant terms on the two sides of the previous equation gives

Sine term :  6B2 ¼ 3

Cosine term :  6B1 ¼ 0
1
It follows that B1 ¼ 0 and B2 ¼  . Thus, the particular solution is
2
1
xp ðtÞ ¼  t cosð3tÞ:
2

The graph of this solution (Figure 2.6.3) clearly shows the amplitude increasing to ∞ with
time, demonstrating an unbounded amplitude of resonant vibration.
88 2 Fundamentals of Mathematics

20

10

x(t ) 0

–10

–20
0 5 10 15 20 25 30 35
Time (seconds)

Figure 2.6.3 Resonant vibration of the system in Example 2.6.11

Example 2.6.12 Deriving a particular solution of a second-order linear differential

equation with constant coefficients and complex roots, using the principle of
superposition
Consider the differential equation

_ þ 14xðtÞ ¼ 3ð1  e5t Þ

x€ðtÞ þ 2xðtÞ

The characteristic equation of the complementary homogeneous equation is

λ2 þ 2λ þ 14 ¼ 0
pffiffiffiffiffi
with complex roots 1  j 13. According to Table 2.6.2 and the principle of superposition,

xp ðtÞ ¼ d0 þ Ae5t :

Substitution of the above form into the differential equation gives

     
25Ae5t þ 2 5Ae5t þ 14 d0 þ Ae5t ¼ 3 1  e5t

which is reduced to
14d0 þ ð25  10 þ 14ÞAe5t ¼ 3  3e5t :

Compare the coefficients of the relevant terms on the two sides of the previous equation to
obtain
 2.6 Solution of Linear Ordinary Differential Equations 89

Constant term : 14d0 ¼ 3

Exponential term : 29A ¼ 3

which yields d0 ¼ 3=14 and A ¼ 3=29. Hence, the particular solution is

3 3
xp ðtÞ ¼  e5t :
14 29

Now, consider a general forcing function

f ðtÞ ¼ eat ½Ak ðtÞcos ωt þ Bm ðtÞsin ωt (2.6.41)

where Ak ðtÞ and Bm ðtÞ are polynomials of orders k and m, respectively. As can be seen, the
three forms of forcing functions in Table 2.6.2 are special cases of Eq. (2.6.41). Define a
complex number
pffiffiffiffiffiffiffi
γ ¼ α þ jω; j ¼ 1 (2.6.42)

where α and ω are the parameters in Eq. (2.6.41) of the forcing function. A particular
solution of the differential equation (2.6.31) is

xp ðtÞ ¼ tμ eat ½Qn ðtÞcos ωt þ Rn ðtÞsin ωt (2.6.43)

where Qn ðtÞ and Rn ðtÞ are polynomials of order n ¼ maxðk; mÞ, as given in Eq. (2.6.29), and μ
is an integer that depends on the value of γ as below

Case 1: If γ ≠ λ1 and γ ≠ λ2 , then μ ¼ 0.

Case 2: If γ ¼ λ1 or γ ¼ λ2 but λ1 ≠ λ2 , then μ ¼ 1.
Case 3: If γ ¼ λ1 ¼ λ2 , then μ ¼ 2.

Example 2.6.13 Deriving a particular solution of a second-order linear differential

equation with constant coefficients and real distinct roots, where the parameter of a
forcing function coincides with one of the roots
For the differential equation

_ þ 12xðtÞ ¼ ð1 þ 2tÞe3t
x€ðtÞ þ 7xðtÞ

the characteristic equation of the complementary homogeneous equation is

λ2 þ 7λ þ 12 ¼ 0
90 2 Fundamentals of Mathematics

with roots 3 and 4. By Eq. (2.6.43), the parameter γ ¼ 3, which coincides with one of the
characteristic roots. By Case 2 of Eq. (2.6.43), a particular solution needs to be used, of the form

xp ðtÞ ¼ te3t ðat þ bÞ ¼ e3t ðat2 þ btÞ

Substituting the above expression into the differential equation and using Eq. (2.6.40) yields
 
9e3t at2 þ bt  6e3t ð2at þ bÞ þ 2ae3t
   
þ 7 3e3t at2 þ bt þ e3t ð2at þ bÞg þ 12e3t at2 þ bt ¼ ð1 þ 2tÞe3t

which is simplified as
2at þ ð2a þ bÞge3t ¼ ð1 þ 2tÞe3t

or
2a þ b þ 2at ¼ 1 þ 2t

Comparison of the t power terms on the two sides of the previous equation gives

t1 term : 2a ¼ 2
t0 term : 2a þ b ¼ 1

which leads to a ¼ 1 and b ¼ 1. So, the particular solution is

xp ðtÞ ¼ te3t ðt  1Þ

General Differential Equations

Now consider the n-th order differential equation (2.6.17). Given forcing function f ðtÞ of the
form of Eq. (2.6.41), define a complex number γ ¼ α þ jω as shown in Eq. (2.6.42). Let the
roots of the characteristic equation (2.6.22) be λ1 ; λ2 ; . . . ; λn . A particular solution of Eq.
(2.6.17) is guessed as belonging to one of the following two cases.

Case 1. The parameter γ does not match any of the roots: γ ≠ λj for j ¼ 1; 2; . . . ; n.
A particular solution can be written as

xp ðtÞ ¼ eat ½Qn ðtÞcos ωt þ Rn ðtÞsin ωt (2.6.44)

Case 2. The parameter γ is the same as a root of multiplicity m (m identical roots). A

particular solution is

xp ðtÞ ¼ tm eαt ½Qn ðtÞcos ωt þ Rn ðtÞsin ωt (2.6.45)

In the above equations Qn ðtÞ and Rn ðtÞ are polynomials of order n ¼ maxðk; mÞ.
 2.6 Solution of Linear Ordinary Differential Equations 91

Example 2.6.14 Deriving a particular solution of a higher-order linear differential

equation with constant coefficients
Consider the following differential equation

€x˙ðtÞ þ 5€ _ þ 3xðtÞ ¼ 2tet

x ðtÞ þ 7xðtÞ

From Example 2.6.8, the characteristic equation has roots 1;  1; and  3. Because the
parameter γ ¼ 1, a particular solution by Eq. (2.6.45) is

xp ðtÞ ¼ t2 et ðat þ bÞ

Substituting the above form into the differential equation gives

 
xpnþ et 9at2 þ 6bt  2b  12at  4boþ 6a  
 
þ 5 xp þ et 6at2 þ 4bt þ 6at þ 2b þ 7 xp þ et ð3at2 þ 2btÞ þ 3xp ¼ 2tet

which is reduced to
 
et 9at2 þ 6bt  6at  2b  12at
  4b þ 6a
þ 5et 6at2  4bt þ 6at þ 2b þ 7et ð3at2 þ 2btÞ ¼ 2tet

This eventually leads to

12at þ ð6a þ 4bÞ ¼ 2t

Comparing the coefficients of the t power terms on the two sides of the previous equation gives the
coupled algebraic equations

12a ¼ 2
6a þ 4b ¼ 0

which have the solutions a ¼ 1=6 and b ¼ 1=4. So, the particular solution of the given differen-
tial equation is

1 t 2
xp ðtÞ ¼ e t ð2t  3Þ
12

Complete Solution
Once a particular solution is obtained, the coefficients in Eq. (2.6.23) or Eq. (2.6.24) can be
determined through the introduction of the initial conditions of Eq. (2.6.18), For instance,
for the solution of Eq. (2.6.23), the coefficients A1 ; A2 ; . . . ; An can be determined by using the
coupled algebraic equations below
92 2 Fundamentals of Mathematics

8
>
> xð0Þ ¼ A1 þ A2 þ . . . þ An þ xp ð0Þ ¼ a0;0
<
Dxð0Þ ¼ A1 λ1 þ A2 λ2 þ . . . þ An λn þ Dxp ð0Þ ¼ a0;1
(2.6.46)
>
> ......
: n1
D xð0Þ ¼ A1 λ1 n1 þ A2 λ2 n1 þ . . . þ An λn n1 þ Dn1 xp ð0Þ ¼ a0;n1

where D ¼ d=dt, and xp ðtÞ is any particular solution of Eq. (2.6.17). For the coefficients
A1 ; A2 ; . . . ; Am and B0 ; B1 ; . . . ; Bp , a set of algebraic equations can be similarly set up
through the application of the initial conditions of Eq. (2.6.18). The determination of the
coefficients in xh ðtÞ eventually yields the complete solution that satisfies both the differential
equation (2.6.17) and the initial conditions of Eq. (2.6.18).

Example 2.6.15 Finding a complete solution of the given linear differential equation
with constant coefficients
Consider the differential equation

_ þ 14xðtÞ ¼ 3ð1  e5t Þ

x€ðtÞ þ 2xðtÞ
xð0Þ ¼ 1; x_ ð0Þ ¼ 0
pffiffiffiffiffi
From Example 2.6.11, the characteristic roots are 1  j 13. This means that the solution of
the complementary homogeneous equation is
 pffiffiffiffiffi  pffiffiffiffiffi 
xh ðtÞ ¼ et B1 sin 13 t þ B2 cos 13 t

Also from Example 2.6.11, a particular solution is

3 3
xp ðtÞ ¼  e5t
14 29

Thus,
 pffiffiffiffiffi  pffiffiffiffiffi  3 3
xðtÞ ¼ xh ðtÞ þ xp ðtÞ ¼ et B1 sin 13 t þ B2 cos 13 t þ  e5t
14 29

From Eq. (2.6.43)

3 3
xð0Þ ¼ 1 : B2 þ þ ¼1
14 29
pffiffiffiffiffi 15
_
xð0Þ ¼0: 13 B1  B2 þ ¼0
29
1 151 361
which have the solutions B1 ¼ pffiffiffiffiffi and B2 ¼ . Therefore, the solution of the given
13 406 406
differential equation is

t 1 151 pffiffiffiffiffi 361 pffiffiffiffiffi 3 3
xðtÞ ¼ e pffiffiffiffiffi sinð 13 tÞ þ cosð 13 tÞ þ  e5t :
13 406 406 14 29
 2.6 Solution of Linear Ordinary Differential Equations 93

2.6.6 Solution of Differential Equations with Piecewise-Continuous Forcing Functions

In this section, differential equations with continuous piecewise forcing functions, as shown
in Section 2.4.6, are considered. We start with an example:

_ þ 4xðtÞ ¼ 3uðtÞ  3uðt  2Þ for t > 0; subject to xð0Þ ¼ 1

xðtÞ

where uðtÞ is the unit step function, uðt  2Þ is a delayed unit step function, and the
forcing function 3uðtÞ  3uðt  2Þ on the right-hand side of the differential equation
represents a pulse as described in Example 2.4.14 and shown in Figure 2.4.2(a). By super-
position, the solution of the above differential equation with an initial condition can be
written as

xðtÞ ¼ xI ðtÞ þ xp1 ðtÞ  xp2 ðtÞ

where xI ðtÞ; xp1 ðtÞ; xp2 ðtÞ are the solutions of the following three problems:

Problem 0 : x_ I ðtÞ þ 4xI ðtÞ ¼ 0; t≥0; subject to xI ð0Þ ¼ 1

Problem1 : x_ p1 ðtÞ þ 4xp1 ðtÞ ¼ 3uðtÞ; t≥0; subject to xp1 ð0Þ ¼ 0
Problem 2 : x_ p2 ðtÞ þ 4xp2 ðtÞ ¼ 3uðt  2Þ; t > 2; subject to xp1 ð2Þ ¼ 0

The solution of Problem 0 (free response) is

xI ðtÞ ¼ e4t

The solution of Problem 1 can be obtained by either the Laplace transform or the method of
undermined coefficients, and it is

3 
xp1 ðtÞ ¼ 1  e4t
4

Likewise, the solution of Problem 2 can be derived as

3 
xp2 ðtÞ ¼ 1  e4ðt2Þ uðt  2Þ
4

It is seen that xp2 ðtÞ is a time-shifted function, derived from xp1 ðtÞ as

xp2 ðtÞ ¼ xp1 ðt  2Þuðt  2Þ

Finally, the solution of the differential equation is

3  3 
xðtÞ ¼ e4t þ 1  e4t þ 1  e4ðt2Þ uðt  2Þ
4 4
94 2 Fundamentals of Mathematics

which can also be written as

8
> 3 
< e4t þ 1  e4t for 0 ≤ t ≤ 2
xðtÞ ¼ 4  
>  
: e4t þ 3 1  e4t þ 3 1  e4ðt2Þ for t > 2
4 4

We now consider the general differential equation

AðDÞxðtÞ ¼ f ðtÞ (2.6.47)

subject to the initial conditions

xð0Þ ¼ a0;0 ; Dxð0Þ ¼ a0;1 ; . . . ; Dn1 xð0Þ ¼ a0;n1 (2.6.48)

where AðDÞ ¼ an Dn þ an1 Dn1 þ . . . þ a1 D þ a0 , with D ¼ d=dt, and the forcing function is

f ðtÞ ¼ f0 ðtÞuðtÞ þ f1 ðt  T1 Þuðt  T1 Þ

(2.6.49)
þ f2 ðt  T2 Þuðt  T2 Þ þ . . . þ fm ðt  Tm Þuðt  Tm Þ

with fj ðt  Tj Þ being a time-shifted function and uðt  Tj Þ being a delayed unit step function.
The solution of the differential equation (2.6.47) can be written as

xðtÞ ¼ xI ðtÞ þ x0 ðtÞuðtÞ þ x1 ðt  T1 Þuðt  T1 Þ

(2.6.50)
þ x2 ðt  T2 Þuðt  T2 Þ þ . . . þ xm ðt  Tm Þuðt  Tm Þ

where xI ðtÞ is the free response that is the solution of the homogeneous equation

AðDÞxðtÞ ¼ 0 (2.6.51)

subject to the initial conditions of Eq (2.6.48), and xj ðtÞ for j ¼ 0; 1; 2; . . . ; m is the

solution of the equation

AðDÞxðtÞ ¼ fj ðtÞ; t > 0 (2.6.52)

subject to zero initial conditions

xð0Þ ¼ 0; Dxð0Þ ¼ 0; . . . ; Dn1 xð0Þ ¼ 0 (2.6.53)

Note that xj ðt  Tj Þ in Eq. (2.6.50) is a time-shifted function derived from xj ðtÞ:

The solution process thus takes the following steps:

(i) Determine the free response xI ðtÞ.

(ii) Determine the forced responses x0 ðtÞ; x1 ðtÞ; . . . ; xm ðtÞ by solving the differential equa-
tion (2.6.52) for j ¼ 0; 1; 2; . . . ; m. Because no time delays are considered, standard
 2.7 Transfer Functions and Block Diagrams of Time-Invariant Systems 95

methods such as the Laplace transform and the method of undetermined coefficients
can be applied.
(iii) Time-shift the solutions obtained in Step (ii): x1 ðt  T1 Þ; x2 ðt  T2 Þ; . . . ; xm ðt  Tm Þ.
(iv) Finally, obtain the complete solution using Eq. (2.6.50).

2.7 Transfer Functions and Block Diagrams of Time-Invariant Systems

Time-invariant dynamic systems are commonly seen in engineering applications. The
spring–mass–damper system in Figure 2.5.1 is one example of such a system. A time-
invariant system is the one whose physical parameters do not change with time. Time-
invariant dynamic systems are governed by differential equations with constant coefficients.
In modeling and analysis of time-invariant systems, three types of models are often used: (a)
differential equations that are derived based on the second law of Newton (Newton’s first law
or law of inertia, that is sometimes called “the first law of motion,” can be viewed as a special
case of the second law with vanishing acceleration or linear momentum); (b) transfer
functions that describe the output-to-input ratio in the Laplace transform domain; and (c)
block diagrams that facilitate the description of complex dynamic systems with multiple
subsystems or components. The derivation of differential equations for a variety of physical
systems will be discussed in the subsequent chapters. In this section we introduce basic
concepts of transfer functions and block diagrams; the application of these concepts to
various dynamic systems are described in Chapters 3 to 9.

2.7.1 Transfer Functions

The transfer function of a time-invariant system is defined as the ratio of the system output to
the system input in the Laplace transform domain with zero initial conditions:


L½output


GðsÞ ¼
(2.7.1)
L½input

zero initial conditions

where L is the Laplace transform operator, and zero initial conditions mean all the initial
values of the input and output and their time derivatives are set to zero.
Assume that a time-invariant system is described by the differential equation (2.6.1), with
rðtÞ and yðtÞ being the input and output of the system, respectively. The Laplace transform of
the differential equation with zero initial conditions gives

AðsÞYðsÞ ¼ BðsÞRðsÞ (2.7.2)

96 2 Fundamentals of Mathematics

where Eq. (2.6.12) with IðsÞ ¼ 0 has been used, and AðsÞ and BðsÞ are the polynomials of s
given in Eq. (2.6.13). By the definition in Eq. (2.7.1), the transfer function of the system
is

YðsÞ BðsÞ bm sm þ bm1 sm1 þ . . . þ b1 s þ b0

GðsÞ ¼ ¼ ¼ (2.7.3)
RðsÞ AðsÞ an sn þ an1 sn1 þ . . . þ a1 s þ a0

As can be seen from Eq. (2.7.3) above, the transfer function does not depend on a
particular input; it depends on the coefficients of the differential equation, which are the
physical parameters of the system. Furthermore, the transfer function given in Eq. (2.7.3)
gives an input-to-output representation as follows:

YðsÞ ¼ GðsÞRðsÞ (2.7.4)

The transfer function given in Eq. (2.7.3) is called an n-th-order transfer function
because the denominator AðsÞ is an n-th-order polynomial of s. In the analysis of dynamic
systems, the information on the poles and zeros of the system transfer functions is often acquired.
The definition and determination of the poles and zeros of an s-domain function have been
discussed in Section 2.4.2, and can be directly applied to transfer functions without further
explanation.
Because AðsÞ and BðsÞ can be obtained from the operator polynomials AðDÞ and BðDÞ,
with D replaced by s, the transfer function given in Eq. (2.7.3) can be derived as follows:


YðsÞ BðDÞ


GðsÞ ¼ ¼
(2.7.5)
RðsÞ AðDÞ

D¼s

Take the spring–mass–damper system in Example 2.5.1, which has the governing differ-
ential equation
 
mD2 þ cD þ k xðtÞ ¼ f ðtÞ (2.7.6)

Assume that the system output and system input are the displacement xðtÞ of the mass and
the external force f ðtÞ, respectively. By Eq. (2.7.4), the transfer function of the spring–mass–
damper system is

X ðsÞ

1
1
GðsÞ ¼ ¼
¼ (2.7.7)
FðsÞ mD2 þ cD þ k D¼s ms2 þ cs þ k

The transfer function of a dynamic system is not unique; depending on the definition
of system input and system output, the same differential equation can have different
transfer functions. For instance, for the previously discussed spring–mass–damper
 2.7 Transfer Functions and Block Diagrams of Time-Invariant Systems 97

_
system, if the output is defined as the velocity xðtÞ of the mass, the transfer function
becomes

_
L½xðtÞ sX ðsÞ s
G1 ðsÞ ¼ ¼ ¼ 2 (2.7.8)
L½f ðtÞ FðsÞ ms þ cs þ k

where Eqs. (2.7.1) and (2.7.7) have been used.

2.7.2 Block Diagrams

Block diagrams are widely used for the representation of dynamic systems in control
engineering. For linear time-invariant systems, block diagrams can be viewed as a diagram-
matic extension of transfer function models. For example, for the spring–mass–damper
system described by the transfer function in Eq. (2.7.7), the block diagram in Figure 2.7.1
relates the input F ðsÞ to the output X ðsÞ by

1
X ðsÞ ¼ FðsÞ
ms2 þ cs þ k

As can be seen from Figure 2.7.1, a block diagram consists of three parts: an arrow
pointing to the block representing the input, an arrow departing from the block
representing the output, and the block (box) containing the transfer function of the
system in it. For general linear time-invariant systems modeled by the transfer function
in Eq. (2.7.3), the block diagram is shown in Figure 2.7.2. The algebraic rule of a block
diagram is

Output ¼ Transfer functionðinside the blockÞ  Input (2.7.9)

Figure 2.7.1 Block diagram of a spring–mass–damper system

Figure 2.7.2 Block diagram of a generic linear time-invariant system

98 2 Fundamentals of Mathematics

Example 2.7.1 Deriving governing equations from the given block diagram
For a dynamic system modeled by the block diagram in Figure 2.7.3, derive the governing
differential equation.
Solution
According to the rule of Eq. (2.7.9) for a block diagram, the output YðsÞ of the system is given by

2s þ 1
YðsÞ ¼ RðsÞ
s2 þ 3s þ 6

or
YðsÞ BðsÞ
¼ ; with BðsÞ ¼ 2s þ 1 and AðsÞ ¼ s2 þ 3s þ 6
RðsÞ AðsÞ

This implies that, in the time domain,

AðDÞyðtÞ ¼ BðDÞrðtÞ; D ¼ d=dt

It follows that the governing differential equation of the system is

y€ðtÞ þ 3yðtÞ
_ þ 6yðtÞ ¼ 2_r ðtÞ þ rðtÞ

Figure 2.7.3 Block diagram of a dynamic system

Block diagrams can be used to describe the input–output relations for systems with
multiple inputs and multiple outputs. Also, block diagrams are the basic element for model-
ing complex dynamic systems with multiple subsystems. Furthermore, block diagrams go
beyond time-invariant systems; the cause-and-effect representation is extended to time--
varying systems and nonlinear systems. These modeling issues will be addressed in the
subsequent chapters.

2.8 Solution of State Equations via Numerical Integration

Many nonlinear dynamic systems and complex linear systems can be modeled by state
equations such as Eq. (2.5.11). Unlike linear time-invariant systems, the analytical methods
presented in Section 2.6 are not valid for the solution of nonlinear state equations. In general,
exact solutions of nonlinear state equations are difficult and often impossible to obtain.
 2.8 Solution of State Equations via Numerical Integration 99

In this section a numerical integration algorithm is presented for deriving approximate

solutions of nonlinear state equations. To this end, write Eq. (2.5.11) as
 
_ ¼ Φ t; xðtÞ
xðtÞ (2.8.1)
   
where Φ t; xðtÞ ¼ ϕ xðtÞ; uðtÞ , with the vector ϕ given in Eq. (2.5.12). The initial condi-
tion of the state equation can be written as

xð0Þ ¼ ϕ0 (2.8.2)

where ϕ0 is a given constant vector. The nonlinear equation (2.8.1) can be solved by a fixed-
step Runge–Kutta method of order four:

h
xkþ1 ¼ xk þ ðf 1 þ 2f 2 þ 2f 3 þ f 4 Þ; k ¼ 1; 2; . . . (2.8.3)
6

where xk ¼ xðtk Þ with tk ¼ kh, h is the step size, and

8
>
> f ¼ Φðt k ; xk Þ
> 1
>
>
> h h
<f 2 ¼ Φ tk þ ; xk þ f 1
 2 2
(2.8.4)
>
> h h
>f 3 ¼ Φ tk þ ; xk þ f 2
>
>
> 2 2
:
f 4 ¼ Φðtk þ h; xk þ hf 3 Þ

The above Runge–Kutta algorithm is a method of numerical integration for differential

equations. In computation, make sure that the step size h is small enough to guarantee the
convergence of the numerical solutions. In advanced numerical integration algorithms, the
step size h is adaptively adjusted during the computation, meaning that “if the solution
appears to be varying rapidly in a particular region, . . . the step size will be reduced to be able
to better track the solution” (Advanced Numerical Differential Equation Solving in
Mathematica, Wolfram Research Institute, 2013). For instance, the MATLAB software
has Runge–Kutta solvers like ode45 with automatic step-size adjustment capability, and
Wolfram Mathematica uses a NDSolve framework that is also based on explicit and implicit
Runge–Kutta methods.
For numerical simulation, the MATLAB commands that implement the algorithm of Eq.
(2.8.3) are given as follows:
h = tf/(npts – 1); % step size
t1 = linspace(0, tf, npts); % time intervals
x = zeros(n, npts); % n = dimension of the state vector
x(:, 1) = x0; % initial values
for ii = 1:npts-1
100 2 Fundamentals of Mathematics

tt = t1(ii); z = x(:,ii);
f1 = feval(FunName, tt, z);
f2 = feval(FunName, tt+h/2, z+h/2*f1);
f3 = feval(FunName, tt+h/2, x+h/2*f2);
f4 = feval(FunName, tt+h, x+h*f3);
x(:,ii+1) = z + h*(f1 + 2*f2 + 2*f3 + f4)/6;
end
 
where feval is a MATLAB function, FunName is a user-provided function for Φ t; xðtÞ , and
npts is the number of points in the region of integration, 0 ≤ t ≤ tf with tf being the total
simulation time. In other words, the step size h of the numerical integration can be expressed
as

tf
h¼ (2.8.5)
npts  1

In the context of Wolfram Mathematica, the NDSolve functionality built-in options can be
used to specify other than an automatically selected numerical method, absolute and relative
errors allowed in the solution, and the maximum number of steps to be taken in attempting to
find a solution. The last option is important when dealing with solutions that have a singular-
ity: if an adaptive step size is chosen, the software will try to reduce the step size many times to
be able to better track the solution approaching singularity, and will get into an infinite loop.
NDSolve can find a solution for a given nonlinear differential equation directly; prior
derivation of the state equations is not necessary. The syntax of a script that finds a solution
of the given differential equation and plots it is as follows:
NDSolve [{Equation, Initial_Conditions}, x, {t, tmin, tmax}]
Plot [ Evaluate [x[t]/.%], {t, tmin, tmax}]

In this script, the first line finds a solution of the given differential equation Equation,
which is defined in terms of the function x(t) and subject to initial conditions
Initial_Conditions; the solution is found for the independent variable t belonging to
the interval tmin ≤ t ≤ tmax . The second line of the script plots the obtained solution.
The example below illustrates the utility of advanced computational software packages
such as MATLAB and Mathematica for finding a numerical solution of a nonlinear differ-
ential equation.

Example 2.8.1 Deriving the trajectory of a ball thrown at an elevation with an initial
velocity
A ball of the mass of 0.1 kg is thrown at an elevation of 1.8 m with an initial velocity shown in
~
Figurep2.8.1. ffi aerodynamic drag acting on the rock is F D ¼  0:0002~
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiThe vj (newtons), where
vj~
vj ¼ x_ þ y_ is the magnitude of the velocity vector. Consider gravity with the gravitational
j~ 2 2
 2.8 Solution of State Equations via Numerical Integration 101

acceleration g = 9.81 m/s. Determine the trajectory of the ball flying in the space for the time
period 0 ≤ t ≤ 5 s.
The equations of motion of the ball are derived using Newton’s second law, such as
( pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
x ¼ 0:0002x_ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
m€ x_ 2 þ y_ 2ffi
y ¼ 0:0002y_ x_ 2 þ y_ 2  mg
m€

which can be cast into the state form of Eq. (2.8.1), with
0 1
0 1 0 1 x3
x x1 B x4 C
B y C B x2 C   B B qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
C
C
xðtÞ ¼ B C ¼ B C; Φ t; xðtÞ ¼ B
@x_ A @x3 A B 0:0002x3 x3 þ x4 =m C
2 2 C
B C
y_ x4 @ qffiffiffiffiffiffiffiffiffiffiffiffiffiffi A
0:0002x4 x3 þ x4 =m  g
2 2

In applying the Runge–Kutta integration algorithm (2.8.3), call the MATLAB function feval
(force_ball, tt, z), where the function force_ball is given as follows.

function f = force_ball(tt, z)
m = 0.1; g = 9.81; v = sqrt(vx^2 + vy^2);
f(1) = z(3); f(2) = z(4);
f(3) = -0.0002* x(3)*v/m;
f(4)= -0.0002* x(4) *v/m – g;

By using the above-mentioned MATLAB commands the trajectory of the ball flying in the two-
dimensional space is obtained in Figure 2.8.2.
To use the Wolfram Mathematica NDSolve functionality, derivation of the state equations is
not necessary; the obtained equations of motion can be used directly as follows (automatic
selection of numerical method, step size, precision, and maximum number of steps is assumed):
(* setting up the numerics *)
m = 0.1; g = 9.81;
vxinit = 35*Cos[47 Degree]; vyinit = 35*Sin[47 Degree];
(* solving the system of differential equations *)
NDSolve
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
0
fm  x″ ½t þ 0:0002  x ½t  x0 ½t2 þ y0 ½t2 ¼¼ 0;
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
″ 0
m  y ½t þ 0:0002  y ½t  x0 ½t2 þ y0 ½t2 þ m  g ¼¼ 0;
0 0
x½0 ¼¼ 0; y½0 ¼¼ 1:8; x ½0 ¼¼ vxinit; y ½0 ¼¼ vyinitg;
fx; yg; ft; 0; 5g;
102 2 Fundamentals of Mathematics

(* plotting the ball trajectory in x-y coordinate plane *)

ParametricPlot[ Evaluate[{x[t], y[t]}/.%], {t,0,5}]

Figure 2.8.1 Problem settings: ball thrown at an elevation with an initial velocity

Ball trajectory in two-dimensional space

35

30

25

20
y (t )

15

10

0
0 10 20 30 40 50 60 70 80 90 100 110
x (t )

Figure 2.8.2 Derived trajectory of the ball in two-dimensional space

As expected, the solutions derived by both software packages are identical.

CHAPTER SUMMARY
The main objective of this chapter is to provide the reader with a review of mathematical
theory and methods instrumental for modeling dynamic systems.
Upon completion of this chapter, you should be able to:
 Problems 103

(1) Apply fundamental principles of vector and matrix algebra to equation solving.
(2) Understand the terminology of complex-number operations and be able to perform
algebraic and trigonometric manipulations on complex numbers.
(3) Use the Laplace transform and its properties for solving linear ordinary differential
equations, commonly encountered in modeling and analysis of dynamic systems.
(4) Use the separation of variables, direct integration, and the method of undetermined
coefficients for solving linear ordinary differential equations with constant coefficients,
which represent linear time-invariant dynamic systems.
(5) Refresh the knowledge of three different methods for modeling dynamic systems: state-
space representation, transfer functions, and block diagrams.
(6) Refresh the knowledge of fundamental principles of solving state equations with numer-
ical integration, using computational software such as MATLAB and Wolfram
Mathematica.

REFERENCES
1. S. W. Goode and S. A. Annin, Differential Equations & Linear Algebra, 4th ed., Pearson, 2015.
2. G. Strang, Differential Equations and Linear Algebra, Wellesley-Cambridge Press, 2014.
3. J. Polking, A. Boggess, and D. Arnold, Differential Equations: Classic Version, 2nd ed., Pearson, 2017.
4. T. F. Bogart, Jr., Laplace Transforms and Control Systems Theory for Technology, Wiley, 1982.
5. E. Kreyszig, “Advanced Engineering Mathematics”, 10th ed., Wiley, 2011.
6. K. Ogata, System Dynamics, 4th ed., Pearson, 2004 .

PROBLEMS
Section 2.1 Vector Algebra
2.1 Perform the following vector operations:
(a) ~
p ¼ 〈2:3 4:1 〉; ~q ¼ 〈6:4 2:5 〉; ~ p þ~q; ~ p  2~q; ~ p ~
q ; 3~ p ~
q
(b) ~p ¼ 〈7:5 1:4 〉; ~q ¼ 〈3:4 2:5 〉; 3~ p þ~q; ~ p ~ q; ~ p ~q ; 2:1~ p  ð1:3Þ~
q
(c) ~
p ¼ 〈2:3 4:1 3:8 〉; ~ q ¼ 〈6:4 2:5 1:7 〉; ~ p þ~ q; ~ p  2~q; ~ p ~q; 3~p ~
q
(d) ~p ¼ 〈7:5 1:4 2:8 〉; ~ q ¼ 〈3:4 2:5 4:9 〉; ~ r ¼ 〈2:4 2:1 3:6 〉;
~ ~ ~ ~ ~
p þ q; p  q; p  q; 2:1~
3~ p  ð1:5Þ~ r; p  ~
q þ 3:1~ ~ r; ~
q ~ r ~q ~
p
2.2 Find the angle between the given vectors:
(a) ~
p ¼ 〈2:7 4:3 〉; ~
q ¼ 〈6:4 2:5 〉
(b) 1:1~p ¼ 〈7:5 1:4 〉; 1:7~
q ¼ 〈3:4 2:5 〉
(c) ~
p ¼ 〈2:3 4:1 3:8 〉; ~ q ¼ 〈6:4 2:5 1:7 〉
~
(d) p ¼ 〈3:4 2:5 4:9 〉; 2~ q ¼ 〈2:4 2:1 3:6 〉
104 2 Fundamentals of Mathematics

2.3 Find the cross-product of the given vectors and show the result in the unit-vector
decomposition form:
(a) ~
p ¼ 〈2:3 4:1 3:8 〉; ~q ¼ 〈6:4 2:5 1:7 〉
(b) ~
p ¼ 〈3:4 2:5 4:9 〉; ~ q ¼ 〈2:4 2:1 3:6 〉
(c) ~
p ¼ 〈7:2 1:4 2:9 〉; ~ q ¼ 〈6:4 5:3 8:1 〉
p ¼ 〈5:3 3:1 6:9 〉; ~
(d) ~ q ¼ 〈3:7 2:1 1:8 〉

2.4 Without computing the angle between the vectors, find out whether the given vectors are
orthogonal, parallel, or neither:
(a) ~p ¼ 〈2:7 2:0 〉; ~
q ¼ 〈6:0 8:1 〉
(b) ~p ¼ 〈7:5 1:5 〉; ~q ¼ 〈0:5 2:5 〉
(c) ~p ¼ 〈2:5 1:5 〉; ~q ¼ 〈0:5 0:3 〉
(d) ~p ¼ 〈2:4 1:5 〉; ~q ¼ 〈0:6 0:3 〉
(e) ~p ¼ 〈2:4 1:0 0:48 〉; ~ q ¼ 〈6:0 2:5 1:2 〉
(f) ~
p ¼ 〈1:2 2:1 5:0 〉; ~ q ¼ 〈3:6 6:2 15:0 〉
(g) ~p ¼ 〈5:0 1:0 0:5 〉; ~ q ¼ 〈4:0 0:8 0:4 〉
(h) ~p ¼ 〈2:4 1:5 0:2 〉; ~ q ¼ 〈0:6 0:3 9:45 〉

Section 2.2 Matrix Algebra

 
1:2 3:5 3:8 5:1
2.5 For the given matrices: A ¼ ; B¼ perform the following
4:8 0:2 0:6 0:3
operations:
(a) A+B
(b) AT + BT
(c) 2A −0.1B
(d) A ∙ AT
(e) A ∙ B
0 1
1:2 3:5 
3:8 5:1 2:0
2.6 For the given matrices: A ¼ @4:8 0:2 A; B ¼ perform the
0:6 0:3 4:2
5:0 1:0
following operations. If an operation is impossible, explain why.
(a) A+B
(b) AT + B
(c) 2A−0.1BT
(d) A ∙ AT
(e) B ∙ A
(f) A ∙ B
(g) B ∙ AT
 Problems 105

2.7 Find the inverses of the given matrices:

  
1:2 3:5 5:1 2:0 3:0 6:0
A¼ ; B¼ ; C¼
0:5 1:2 0:5 4:2 2:1 4:2

If the inverse does not exist, explain why.

2.8 Find the inverses of the given matrices:
0 1 0 1 0 1
1:0 1:2 3:5 5:1 1:3 2:0 1:0 3:0 6:0
A ¼ @0:7 0:5 1:2 A; B ¼ @10:2 2:6 4:0 A; C ¼ @0:4 1:2 2:0 A
1:4 2:0 2:0 0:5 7:1 4:2 0:7 2:1 4:2

If the inverse does not exist, explain why.

2.9 Solve the following systems of linear equations by Cramer’s rule:
8
<2x  5y  4z ¼ 1
(a) x þ 3y þ 4z ¼ 5
:
3x  8y þ z ¼ 12
8
<x  2y þ z ¼ 1
(b) x þ 3y þ 4z ¼ 8
:
3x  6y þ 3z ¼ 3
8
<x þ 2:1y  3z ¼ 0
(c) 4:2x  3y þ 5:2z ¼ 6:2
:
yþz¼0

Section 2.3 Complex Numbers

2.10 Find the magnitude and phase for the given complex numbers:
(a) z ¼ 2 þ j7
(b) z ¼ 2 þ j5
(c) z ¼  4  j3
(d) z ¼ 5  j9

2.11 Find real and imaginary parts of the given complex numbers:
2 þ j7
(a) z ¼
1þj
2 þ j5
(b) z ¼
1 þ j3
 4  j3
(c) z ¼
2j
5  j9
(d) z ¼
3  4j
106 2 Fundamentals of Mathematics

2.12 Multiply the given complex numbers using Euler’s formula. Find the magnitude and
phase of the multiplication result:
2 þ j7
(a) z ¼  ð3 þ j4Þ
1þj
9 þ j
(b) z ¼ ð4 þ j5Þ 
2  j3
6  j2  4  j3
(c) z ¼ 
5þj 2j

Section 2.4 The Laplace Transform

2.13 Derive the Laplace transform of the following functions:
(a) f ðtÞ ¼ 2 þ e2t
(b) f ðtÞ ¼ 3t  t2
(c) f ðtÞ ¼ 5te4t
(d) f ðtÞ ¼ 6 sinð3tÞ þ 8 cosð5tÞ
(e) f ðtÞ ¼ t sinð2tÞ
(f) f ðtÞ ¼ ðt  3Þe2ðt3Þ uðt  3Þ
(g) f ðtÞ ¼ e2t ðt  4Þ
(h) f ðtÞ ¼ 3 cosð5tÞe6t
(i) f ðtÞ ¼ e3tþ3 sinðt  1Þuðt  1Þ
(j) f ðtÞ ¼ e3tþ5 sinðt  1Þuðt  1Þ
(k) f ðtÞ ¼ 7tetþ2 cosð3tÞ

2.14 Derive the Laplace transform of the following piecewise-continuous functions:

8
<2 t<3
(a) f ðtÞ ¼ 3 3 < t ≤ 5
:
0 t>5
8
< 0 t<3
(b) f ðtÞ ¼ 2 3 < t≤5
:
t4 t>5
8
< 1 t<2
(c) f ðtÞ ¼ t  2 2 < t ≤ 6
:
4 t>6
 Problems 107

2.15 Derive the Laplace transform of the piecewise-continuous functions defined in the
graphs in Figure P2.1:

5
3.0

4 2.5

2.0
3
f (t)

f (t)
1.5
2
1.0
1 0.5

0 0.0
–0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 1 2 3
t t
(a) (b)
3 3

2 2
f (t)

f (t)
1 1

0 0

–1 –1
0 1 2 3 0 1 2 3
t t
(c) (d)
3 5

4
2
3
f (t)
f (t)

1 2

1
0
0

–1 –1
0 1 2 3 0 1 2 3 4 5
t t
(e) (f)

Figure P2.1 Piecewise-continuous functions

2.16 Derive the inverse Laplace transform of the following functions. Use the table of
Laplace pairs and/or apply algebraic operations to the given function to represent it
as a summation of components, the inverse transform of which is known:
4s
(a) FðsÞ ¼
þ9 s2
5
(b) FðsÞ ¼ 2
s þ4
108 2 Fundamentals of Mathematics

5
(c) FðsÞ ¼
s2 þ 2s þ 5
s2  9
(d) FðsÞ ¼
ðs2 þ 9Þ2
3
(e) FðsÞ ¼
ðs þ 1Þðs þ 2Þ
3s þ 4
(f) FðsÞ ¼
ðs þ 1Þðs þ 2Þ
7
(g) FðsÞ ¼
sðs þ 5Þ
2.17 Derive the inverse Laplace transform of the following functions. Use the table of
Laplace pairs and/or apply algebraic operations to the given function to represent it
as a summation of components, the inverse transform of which is known:
8
(a) FðsÞ ¼ 2
s 9
4
(b) FðsÞ ¼ 2
sðs þ 6s þ 13Þ
2
(c) FðsÞ ¼
sðs þ 1Þ2
5s
(d) FðsÞ ¼
ðs þ 9Þ2
2

s2 þ 3s  1
(e) FðsÞ ¼
ðs2 þ 4Þ2
2s þ 3
(f) FðsÞ ¼ 2
s ðs þ 5Þ
sþ6
(g) FðsÞ ¼ 2
s þ 8s þ 20
3s þ 5
(h) FðsÞ ¼ 2
ðs þ 4s þ 21Þðs2 þ 2s þ 5Þ
s2 þ 2s þ 4
(i) FðsÞ ¼ 2
ðs þ 4s þ 22Þðs2 þ 2s þ 14Þ
s2 þ 2s þ 3
(j) FðsÞ ¼ 3
s þ 10s2 þ 31s þ 30

2.18 Derive the inverse Laplace transform of the following functions using the partial
fraction expansion

5
(a) FðsÞ ¼
sðs þ 2Þðs þ 3Þ
3s þ 1
(b) FðsÞ ¼
sðs þ 1Þðs þ 4Þ
 Problems 109

3s þ 2
(c) FðsÞ ¼
s2 ðs þ 1Þðs þ 4Þ
4s
(d) FðsÞ ¼ 3
s ðs þ 2Þ
4s2 þ s þ 7
(e) FðsÞ ¼
s2 ðs þ 2Þ2
s2 þ 4s þ 1
(f) FðsÞ ¼ 4
s þ 6s3 þ 12s2 þ 10s þ 3
2.19 Derive the inverse Laplace transform of the following functions using the partial
fraction expansion
s2 þ 2s þ 2
(a) FðsÞ ¼
s3 þ 5s2 þ 17s þ 13
s3 þ 2s2 þ 1
(b) FðsÞ ¼ 4
s þ 4s3 þ 13s2
5s þ 2
(c) FðsÞ ¼ 4
s þ 8s3 þ 25s2 þ 36s þ 20
s2 þ 3s þ 7
(d) FðsÞ ¼ 4
s þ 8s þ 26s2 þ 40s þ 25
3

2 þ e2s
(e) FðsÞ ¼ 3
s þ 11s2 þ 36s þ 26

Section 2.5 Differential Equations

2.20 Determine whether each of the given differential equations is linear or nonlinear, and
provide a rationale for your answer:
(a) 2xx_ þ x ¼ e2t ; x ¼ xðtÞ
(b) x_ þ t2 x ¼ 0; x ¼ xðtÞ
(c) 4€x þ ðt  1Þx_ þ 2x ¼ 0; x ¼ xðtÞ
(d) x€ þ 3x_ þ cosðxÞ ¼ 0; x ¼ xðtÞ
(e) e4t x€ þ 3x_  12x ¼ 0; x ¼ xðtÞ

2.21 Write the given equations in D-operator notation:

(a) 5xð5Þ þ 2:1xð3Þ þ x€  4x_ þ 12x ¼ f ð3Þ þ 7f€ þ 2f

(b) xð4Þ þ 0:3xð3Þ þ 1:1€x  x ¼ 2f€  f_ þ f
(c) 7:2x þ 4:3x  3:1xð3Þ þ 2:8€
ð6Þ ð5Þ
x  9:2x_ þ 12x ¼ 4f ð4Þ þ f€  5f_ þ f
110 2 Fundamentals of Mathematics

2.22 Write the given systems of equations in matrix notation:

m1 x€1 þ cx_ 1 þ ðk1 þ k2 Þx1  k2 x2 ¼ f1 ðtÞ

(a)
m2 x€2 þ k2 x2  k2 x1 ¼ f2 ðtÞ
m1 x€1 þ cx_ 1 þ k1 x1  cx_ 2 ¼ f1 ðtÞ
(b)
m2 x€2 þ cx_ 2  cx_ 1 ¼ f2 ðtÞ
m1 x€1 þ ðc1 þ c2 Þx_ 1 þ ðk1 þ k2 Þx1  c2 x_ 2  k2 x2 ¼ f1 ðtÞ
(c)
m2 x€2 þ c2 x_ 2 þ k2 x2  c2 x_ 1  k2 x1 ¼ f2 ðtÞ

Section 2.6 Solution of Linear Differential Equations with Constant Coefficients

2.23 Solve the following first-order linear differential equations by direct integration:
(a) 3x_ ¼ e2t ; xð0Þ ¼ 1
(b) 2x_ ¼ et þ 1; xð0Þ ¼ 1
(c) 4x_ ¼ t2 ; xð0Þ ¼ 2
(d) x_ ¼ ðt þ 2Þ2 ; xð0Þ ¼ 4
(e) 5x_ ¼ sinð2tÞ; xð0Þ ¼ 5
(f) 8x_ ¼ t cosðt2 Þ; xð0Þ ¼ 2

2.24 Solve the following second-order linear differential equations by direct integration or
the separation of variables:
(a) 4€x ¼ e3t ; xð0Þ ¼ 1; xð0Þ
_ ¼1
(b) 2€x ¼ e5t þ 3t þ t2 ; xð0Þ ¼ 1; xð0Þ
_ ¼2
2 _
(c) x€ ¼ 2t  1; xð0Þ ¼ 2; xð0Þ ¼ 1
(d) 5€x ¼ ð10t þ 3Þ2 ; xð0Þ ¼ 6; xð0Þ
_ ¼8

2.25 Solve the following differential equations by the separation of variables:

3
(a) 2x_ ¼ e3t ; xð0Þ ¼ 3
x
(b) x_ þ 2x ¼ 5; xð0Þ ¼ 1
(c) 3x_ ¼ x  e2t ; xð0Þ ¼ 2
(d) 2x_ ¼ sinð2xÞ; xð0Þ ¼ π=4
(e) x_ ¼ x2 cosð2tÞ; xð0Þ ¼ π=3
(f) x_ ¼ t e2x ; xð0Þ ¼ 1

2.26 Solve the following first-order differential equations by the Laplace transform:
(a) 2x_ þ x ¼ 4 þ t; xð0Þ ¼ 1
(b) x_ þ 5x ¼ 8t sinð3tÞ; xð0Þ ¼ 1
(c) x_ þ 6x ¼ 3et cosð2tÞ; xð0Þ ¼ 2
(d) x_ þ 3x ¼ 2tet sinð3tÞ; xð0Þ ¼ 37=169
(e) 2x_  x ¼ 4t2 sinðt=2 Þ; xð0Þ ¼ 4
(f) x_  2x ¼ 7t2 e5t þ t cosð2tÞ; xð0Þ ¼ 5=49
 Problems 111

2.27 Solve the following differential equations by the Laplace transform:

(a) 5€x þ 4x_ þ 20x ¼ 20; xð0Þ ¼ 1; xð0Þ _ ¼1
(b) x€ þ 8x_ þ 25x ¼ 25t; xð0Þ ¼ 2; xð0Þ_ ¼ 1
(c) 5€x þ 24x_ þ 180x ¼ 25 t þ 50; xð0Þ ¼ 2; xð0Þ _ ¼0
(d) x€ þ 4x_ þ 16x ¼ 16 t e2t ; xð0Þ ¼ 0; xð0Þ
_ ¼3
(e) _
x€ þ 3x_ þ 9x ¼ 9t þ t2 ; xð0Þ ¼ 0; xð0Þ ¼0
(f) x€ þ 4x_ þ 25x ¼ 25et ; xð0Þ ¼ 1; xð0Þ_ ¼1
(g) x€ þ 2x_ þ 25x ¼ 5 sinðtÞ; xð0Þ ¼ 2; xð0Þ
_ ¼0
(h) x€ þ x_ þ x ¼ cosðtÞ; xð0Þ ¼ 1; xð0Þ
_ ¼0

2.28 Solve the following differential equations by the Laplace transform:

(a) x€ þ 7x_ þ 10x ¼ 5; xð0Þ ¼ 1; xð0Þ_ ¼1
_
(b) x€ þ 6x_ þ 8x ¼ 2t; xð0Þ ¼ 2; xð0Þ ¼ 1
(c) x€ þ 11x_ þ 24x ¼ 12t þ 1; xð0Þ ¼ 2; xð0Þ
_ ¼0
(d) x€ þ 4x_ þ 4x ¼ 2 sinðtÞ; xð0Þ ¼ 0; xð0Þ
_ ¼0
(e) x€ þ 16x_ þ 64x ¼ 169 cosðtÞ; xð0Þ ¼ 1; xð0Þ
_ ¼ 1
(f) x€ þ 4x_ þ x ¼ 3e2t ; xð0Þ ¼ 0; xð0Þ
_ ¼1
(g) x€ þ 8x_ þ 16x ¼ 16 t et ; xð0Þ ¼ 1; xð0Þ
_ ¼2

2.29 Solve the following differential equations with piecewise-continuous forcing functions
and zero initial conditions by the Laplace transform:

1 t<2
(a) x€ þ 2x_ þ 4x ¼ _
; xð0Þ ¼ 0; xð0Þ ¼0
0 t>2
8
<0 t<1
(b) x€ þ 4x_ þ 3x ¼ 2 1 < t < 3 ; xð0Þ ¼ 0; xð0Þ _ ¼0
:
80 t > 3
<0 t<0
(c) x€ þ 8x_ þ 16x ¼ 4 t 0 ≤ t < 1 ; xð0Þ ¼ 0; xð0Þ_ ¼0
:
4 t≥1
8
<0 t<0
(d) x€ þ 9x ¼ t 0 ≤ t < 2 ; xð0Þ ¼ 0; xð0Þ_ ¼0
:
2 t≥2
8
>
> 0 t<0
<
1 0≤t < 1
(e) x€ þ 4x_ þ 29x ¼ _
; xð0Þ ¼ 0; xð0Þ ¼0
>
> 2  t 1≤t < 3
:
1 t≥3

2.30 Solve the following first-order differential equations by the method of undetermined
coefficients:
(a) 2x_ þ 4x ¼ 0; xð0Þ ¼ 1
112 2 Fundamentals of Mathematics

(b) 2x_ þ x ¼ sinðtÞ; xð0Þ ¼ 1

(c) x_ þ 3x ¼ 5 cosð2tÞ; xð0Þ ¼ 2
(d) x_ þ 2x ¼ e2t ; xð0Þ ¼ 5
(e) x_ þ 2x ¼ t2 þ 3; xð0Þ ¼ 1
(f) x_ þ 3x ¼ t2 et ; xð0Þ ¼ 1
(g) 2x_ þ 6x ¼ t e3t ; xð0Þ ¼ 2

2.31 Solve the following homogeneous second-order differential equations by the method of
undetermined coefficients:
(a) x€ þ 5x_ þ 6x ¼ 0; xð0Þ ¼ 1; xð0Þ
_ ¼2
€ _ _
(b) x þ 8x þ 16x ¼ 0; xð0Þ ¼ 1; xð0Þ ¼ 1
(c) x€ þ 4x_ þ 13x ¼ 0; xð0Þ ¼ 1; xð0Þ _ ¼1
(d) x€ þ 6x_ þ 5x ¼ 0; xð0Þ ¼ 5; xð0Þ_ ¼1
(e) x€ þ 6x_ þ 9x ¼ 0; xð0Þ ¼ 4; xð0Þ
_ ¼ 2
(f) x€ þ 2x_ þ 17x ¼ 0; xð0Þ ¼ 2; xð0Þ
_ ¼1

2.32 Solve the following nonhomogeneous second-order differential equations by the

method of undetermined coefficients:
(a) x€ þ 5x_ þ 6x ¼ 3 sinð2tÞ; xð0Þ ¼ 2; xð0Þ
_ ¼0
_
(b) x€ þ 5x_ þ 6x ¼ 2 cosð3tÞ þ t2 ; xð0Þ ¼ 1; xð0Þ ¼ 2
2 2t
(c) x€ þ 5x_ þ 6x ¼ 4t e ; xð0Þ ¼ 3; xð0Þ_ ¼2
_
(d) x€ þ 8x_ þ 16x ¼ t2 þ 5; xð0Þ ¼ 299=128 ; xð0Þ ¼ 81=16
(e) x€ þ 8x_ þ 16x ¼ 3 sinðtÞ; xð0Þ ¼ 61=289 ; xð0Þ
_ ¼ 176=289
4t
(f) x€ þ 8x_ þ 16x ¼ e ðsinðtÞ þ 2Þ; xð0Þ ¼ 5 ; xð0Þ
_ ¼ 23
(g) x€ þ 6x_ þ 5x ¼ 2e5t þ 4tet ; xð0Þ ¼ 27=4; xð0Þ
_ ¼1

2.33 Solve the following nonhomogeneous second-order differential equations by the

method of undetermined coefficients:
(a) x€ þ 4x_ þ 13x ¼ 45e5t cosð3tÞ; xð0Þ ¼ 5; xð0Þ
_ ¼ 1
(b) x€ þ 4x_ þ 13x ¼ 24e2t t sinð3tÞ; xð0Þ ¼ 9; xð0Þ
_ ¼ 3
3t
(c) x€ þ 6x_ þ 9x ¼ 4e þ 27t; xð0Þ ¼ 3; xð0Þ _ ¼ 10
(d) x€ þ 6x_ þ 9x ¼ 36t2 e3t ; xð0Þ ¼ 2; xð0Þ
_ ¼8
(e) x€ þ 9x ¼ 2 sinð3tÞ þ 12 t cosð3tÞ; xð0Þ ¼ 1; xð0Þ_ ¼ 3
t
(f) x€ þ 4x ¼ 17e sinð2tÞ; xð0Þ ¼ 7; xð0Þ
_ ¼ 20
(g) x€ þ 2x_ þ 17x ¼ 24t2 et cosð4tÞ; xð0Þ ¼ 2; xð0Þ
_ ¼6

2.34 Solve the equations of Problem 2.29 by the method of undetermined coefficients as
described in Section 2.6.6.
Section 2.7 Transfer Functions and Block Diagrams of Time-Invariant Systems
 Problems 113

2.35 Derive transfer functions of the following systems:

(a) x€ þ 2x_ þ 13x ¼ f ðtÞ

(b) xð3Þ þ 4€x þ 6x_ þ 2x ¼ f ðtÞ
(c) 2xð3Þ þ 9x_ þ 25x ¼ 4f ðtÞ
(d) 2xð3Þ þ 7x_ þ 3x ¼ 12f ðtÞ  f_ ðtÞ
(e) 5xð4Þ  x€ þ 3x_ þ 8x ¼ 4f_ ðtÞ  7f ðtÞ

2.36 Derive the governing equations for the block diagrams given in Figure P,2,2:

F(s) 3s +1 X(s) F(s) 25 X(s)

s3 + 4s2 + s + 7 s3 + 6s +12

(a) (b)
F(s) 5s 3 X(s) F(s) s+4 X(s)
s2 + 7s +10 s3 + 2s2 + 5s +8

(c) (d)
F(s) s +2
2 X(s) F(s) s s +1
2 X(s)
s5 +8s3 + 7s2 + 6 s4 + 7s2 + 6s +11

(e) (f)

Figure P2.2 Block diagrams

Section 2.8 Solution of State Equations via Numerical Integration

2.37 Solve Problem 2.26 by numerical integration using Mathematica.
2.38 Solve Problem 2.26 by numerical integration using MATLAB.
2.39 Solve Problem 2.30 by numerical integration using Mathematica.
2.40 Solve Problem 2.30 by numerical integration using MATLAB.
2.41 Considering zero initial conditions, solve the system of linear differential equations by
numerical integration using Mathematica and plot the variables x1 and x2 vs. time

x€1 þ 5x_ 1 þ 6x1  5x_ 2 ¼ sinð2tÞ

4€x 2 þ 5x_ 2  5x_ 1 ¼ 2 cosð3tÞ

2.42 Solve Problem 2.41 by numerical integration using MATLAB and plot the variables x1
and x2 vs. time
3 Mechanical Systems

Contents
3.1 Fundamental Principles of Mechanical Systems 115
3.2 Translational Systems 120
3.3 Rotational Systems 144
3.4 Rigid-Body Systems in Plane Motion 169
3.5 Energy Approach 177
3.6 Block Diagrams of Mechanical Systems 181
3.7 State-Space Representations 185
3.8 Dynamic Responses via MATLAB 191
3.9 Dynamic Responses via Mathematica 198
Chapter Summary 204
References 204
Problems 204

Mechanical systems are seen in machines, devices, equipment, and structures in a wide
variety of engineering applications. Modeling of a mechanical system involves forces and
motion about relevant objects, which can be either solids or fluids. Depending on specific
concerns in an application, there are several types of mathematical models for mechanical
systems at different levels of complexity, including lumped-parameter models, rigid-body
models, and deformable- or flexible-body models. For instance, in studying the dynamic
behaviors of an airplane, a lumped-parameter model may be good enough to describe its
flight trajectory and a rigid-body model may be sufficient to study its three-dimensional
motion in space. However, a flexible-body model is necessary to understand the vibration
and stress in the airplane structure in performance evaluation and failure analysis.
Regardless of the level of complexity of the mathematical model to be created, there are
three importing keys in modeling of mechanical systems:

• fundamental principles
• basic elements
• ways of analysis

114
 3.1 Fundamental Principles of Mechanical Systems 115

Fundamental principles (such as Newton’s laws of motion) are the first laws of nature
relating forces to motion. Basic elements (like masses, springs, and dampers) are approxi-
mate models of those components that are the building blocks of mechanical systems. A way
of analysis (free-body diagram for instance) is an approach by which fundamental principles
are applied to basic elements such that a mathematical model is established. The model so
obtained is a set of governing equations of motion, which are differential and algebraic
equations. The three-key modeling concept will be illustrated throughout this chapter.
In this text, the International System of Units (SI) is used. Refer to Tables A1 and A2 in
Appendix A for commonly used quantities for mechanical systems, and Table A5 for unit
conversions between the International System and the US customary system.

3.1 Fundamental Principles of Mechanical Systems

Without loss of generality, we shall first consider particles that are objects with negligible
dimensions and with masses lumped to points. Such particles are a lumped-parameter model
of mass elements of mechanical systems. The fundamental principles of particles shall be
extended to rigid and flexible bodies later in the text.
This section provides a quick review of terms, definitions, and formulas in particle
dynamics, as a reference for modeling of mechanical systems. Detailed explanations of
these contents can be found in standard textbooks on dynamics, such as those listed at the
end of the chapter. Readers who are familiar with these contents can skip this section.

3.1.1 Newton’s Laws of Motion

Newton’s three laws of motion for particles can be stated as follows:

First law: A particle remains at rest or moves with a constant velocity along a straight line if
the sum of the forces acting on it is zero.
Second law: The rate of change of the linear momentum of a particle is equal to the sum of the
forces acting on it.
Third law: The forces exerted by two particles upon each other, which are called the action
and reaction forces, are equal in magnitude, opposite in direction, and collinear.

In the above-mentioned laws of motion, inertia frames of reference are considered. An

inertial frame of reference is one in which Newton’s first law of motion is valid.
Mathematically, Newton’s second law is expressed by

d
F¼ ðmvÞ (3.1.1)
dt
116 3 Mechanical Systems

where m and v are the mass and velocity of the particle, respectively; F is the resultant or sum
of all forces acting on the particle; and mv is the linear momentum of the particle. Newton’s
second law is also applicable to rigid and flexible bodies.
If the mass m is constant, Eq. (3.1.1) becomes

F ¼ ma (3.1.2)

dv
where a ¼ is the acceleration of the particle. Equation (3.1.2) is a commonly used
dt
expression of Newton’s second law, which indicates that the acceleration of a particle of
constant mass is proportional to the resultant force. Also, Newton’s first law can be viewed
as a special example of the second law with vanishing linear momentum or acceleration

F¼0 (3.1.3)

Newton’s laws can be extended to a system of particles. For a system of N particles,

Newton’s second law can be written as
( )
XN
d X N
Fi ¼
ext
mi vi (3.1.4)
i¼1
dt i¼1

i is the resultant external force applied to the i-th particle, and mi and vi are the mass
where F ext
and velocity of the i-th particle. Here, those action and reaction forces between any two
particles are known as internal forces and those forces that are not internal forces are called
external forces. Internal forces do not appear in Eq. (3.1.4) because they cancel each other.

3.1.2 Forces and Moments

According to Newton’s second law, the motion of a particle or body is caused by forces. A
force is an action of pull or push applied by one body to another. There are various types of
forces, including external loads, gravity, friction, and spring forces. A force is characterized by
its magnitude, direction of action, and point of application. Thus, forces can be expressed by
vectors and the superposition of forces follows the parallelogram rule of vector summation,

F ¼ F1 þ F2 (3.1.5)

The moment of a force is a measure of the tendency of the force to rotate a body. The
moment M O of force F about a reference point O is defined by the vector cross-product

MO ¼ r  F (3.1.6)

where r is a position vector that runs from the reference point to the point of application of F;
see Figure 3.1.1. The direction of the moment is determined by the right-hand rule, as shown
in the figure. Refer to Section 2.1.4 for determination of vector cross-products.
 3.1 Fundamental Principles of Mechanical Systems 117

MO
Figure 3.1.1 The
moment of a force
about point O

O
F
r

Figure 3.1.2 Two

coordinate systems:
r(t ) r (t ) (a) Cartesian coord-
inates; (b) cylindrical
k z k z
coordinates
0 j 0
i
x r
θ eθ
y
er
(a) (b)

3.1.3 Kinematics of Particles

When a particle moves along a path in a reference frame, its position is described by a
position vector r(t), which is a function of time. The velocity vðtÞ and acceleration aðtÞ of the
particle are the first and second derivatives of r(t)

drðtÞ d2 rðtÞ dvðtÞ

vðtÞ ¼ ¼ r_ ðtÞ; aðtÞ ¼ ¼ ¼ r€ðtÞ (3.1.7)
dt dt2 dt

The motion of a particle can be described in different coordinates. Two commonly used
coordinate systems are Cartesian coordinates and cylindrical coordinates; see Figure 3.1.2,
where i, j, k and er , eθ , k are unit vectors. Note that the directions of er and eθ change with
angle θ.
In Cartesian coordinates, position, velocity, and acceleration vectors are written as
118 3 Mechanical Systems

rðtÞ ¼ xi þ yj þ zk
_ þ yj
vðtÞ ¼ xi _ þ z_ k (3.1.8)
aðtÞ ¼ x€i þ y€j þ z€k

For plane motion, the expressions of position, velocity, and acceleration vectors can be
obtained from Eq. (3.1.8) by neglecting the coordinate z. In the cylindrical coordinates,
position, velocity, and acceleration vectors are of the form

rðtÞ ¼ rer þ zk
_ θ þ z_ k
vðtÞ ¼ r_ er þ rθe (3.1.9)
 
r  rθ_ Þer þ rθ€ þ 2_r θ_ eθ þ z€k
2
aðtÞ ¼ ð€

The second and third lines of Eq. (3.1.9) are obtained via differentiation of the first one and
through use of the relations

_ θ ; e_ θ ¼ θ_ e_ r
e_ r ¼ θe (3.1.10)

For plane motion, the cylindrical coordinates are reduced to the polar coordinates (r, θ) and
position, velocity, and acceleration vectors become

rðtÞ ¼ rer
_ θ
vðtÞ ¼ r_ er þ rθe (3.1.11)
 
r  rθ_ Þer þ rθ€ þ 2_r θ_ eθ
2
aðtÞ ¼ ð€

3.1.4 Principle of Work and Energy

Consider a particle in motion. Dot-product Eq. (3.1.1) by dr and integrate the resulting
equation between positions r1 and r2 on a path of the particle, to obtain

ðr2 ðt2
1 d
F  dr ¼ m ðv  vÞdt
2 dt
r1 (3.1.12)
t1

t
1 2

2 1 2 1 2
¼ mv
¼ mv2  mv1
2 t1 2 2

where v ¼ jvj, and v1 and v2 are the velocity magnitudes at r1 and r2 (at times t1 and t2 ),
respectively. Recall that the kinetic energy of the particle is

1 1
TðtÞ ¼ mv  v ¼ mv2 (3.1.13)
2 2
 3.1 Fundamental Principles of Mechanical Systems 119

and that the work done by force F on a path from r1 to r2 is

ðr2
W12 ¼ F  dr (3.1.14)
r1

Thus, Eq. (3.1.12) states the principle of work and energy

W12 ¼ T2  T1 ¼ ΔT (3.1.15)

where T1 ¼ Tðt1 Þ and T2 ¼ Tðt2 Þ. Equation (3.1.15) shows that the work done by the force
applied to a particle is equal to the change of the kinetic energy of the particle.
A force F is conservative if its work is independent of any path chosen between two
positions r1 and r2 . The work of a conservative force can be expressed by

W12 ¼ ðV2  V1 Þ ¼ ΔV (3.1.16)

where V ¼ VðrÞ, which, being a scalar function of position vector r, is called potential energy,
and V1 ¼ Vðr1 Þ; V2 ¼ Vðr2 Þ. For example, the gravitational force and the spring force are
both conservative, with the following potential energy functions:

Gravity: V ¼ mgy
1
Spring force: V ¼ kx2
2

where y is a spatial coordinate along an upward axis, and x is the spring elongation. A force F
is non-conservative if its work is dependent upon a path chosen. No potential energy exists for
a non-conservative force. For instance, friction is a non-conservative force.
Let a particle be subject to a conservative force F c and a non-conservative force F nc . The
total work done by the forces can be written as

W12 ¼ W12
nc
 ΔV (3.1.17)

where W12nc
is the work done by the non-conservative force F nc . Substitution of Eq. (3.1.17)
into Eq. (3.1.15) gives the principle of work and energy
nc
W12 ¼ ðT2 þ V2 Þ  ðT1 þ V1 Þ ¼ ΔðT þ VÞ (3.1.18)

Here, the sum T þ V is called the mechanical energy of the particle. If only conservative forces
are applied to the particle ðF nc ¼ 0Þ, the mechanical energy of the particle is conserved:

T2 þ V2 ¼ T1 þ V1 (3.1.19)
120 3 Mechanical Systems

3.1.5 Principle of Impulse and Momentum

The time integration of Newton’s second law, Eq. (3.1.1), leads to the principle of impulse and
momentum
I 12 ¼ L2  L1 (3.1.20)
ðt2
where I 12 ¼ Fdt is the impulse of the force applied to the particle, LðtÞ ¼ mv is the linear
t1
momentum of the particle, L1 ¼ Lðt1 Þ and L2 ¼ Lðt2 Þ. If no force is applied to the particle,
which means I 12 ¼ 0, by Eq. (3.1.20), the impulse is conserved:

L2 ¼ L 1 (3.1.21)

For a system of N particles ðm1 ; m2 ; . . . ; mN Þ, the principle of impulse and momentum has
the same form of Eq. (3.1.20). However, by Eq. (3.1.4), the impulse and momentum of the
system are given as follows:

ðt2 X
N X
N
I 12 ¼ Fext
i dt; LðtÞ ¼ m i vi (3.1.22)
i¼1 i¼1
t1

3.2 Translational Systems

A translational system is a mechanical system with its components only involved in transla-
tional motion. Modeling of translational systems usually results in lumped-parameter
models.

3.2.1 Basic Elements

As the first key in modeling of mechanical systems, fundamental principles of particles have
been presented in Section 3.1. As the second key in system modeling, three types of basic
elements are introduced in this section: mass elements, spring elements, and damping
elements.

Lumped Masses
In a lumped-parameter model, lumped masses are used. A lumped mass, which is also called
a discrete mass or point mass or simply mass, is a particle of concentrated inertia, as
described in Section 3.1. The principles and formulas given in Section 3.1 thus are all
applicable to lumped masses. For example, for a mass m in Figure 3.2.1(a), which moves
 3.2 Translational Systems 121

Figure 3.2.1 Lumped masses: (a) horizontal motion; (b) vertical motion

on a smooth horizontal surface and is subject to a force F(t), its displacement x(t) by
Newton’s second law, Eq. (3.1.2), is governed by the differential equation

x¼F
m€

In the figure, the two wheels under the moving mass indicate that the supporting horizontal
surface is smooth without friction. Similarly, for a mass M in Figure 3.2.1(b), which moves
vertically under the influence of gravity and a force F(t), its displacement y(t) is governed by
the differential equation

M y€ ¼ F  Mg

where g is the gravitational acceleration.

Translational Springs
A translational spring or linear spring is a spring of negligible inertia, with its internal force
(spring force) given by Hooke’s law

Spring force ¼ k  Δx (3.2.1)

where k is a spring coefficient, in N/m, and Δx is the elongation of the spring. The Δx is also
the relative displacement of the two ends of the spring. Figure 3.2.2 shows three cases of
springs in deformation: (a) a spring with a fixed end; (b) a spring of two movable ends in
tension; and (c) a spring of two movable ends in compression. In case (a), the spring force fs
and the displacement (elongation) x are in the relation

fs ¼ kΔx ¼ kx (3.2.2)

where the relative displacement Δx ¼ x  0 ¼ x. In case (b), the spring force of the tensioned
spring is given by
122 3 Mechanical Systems

Figure 3.2.2 Springs in deformation: (a) spring with a fixed end; (b) spring of two movable ends in tension;
(c) spring of two movable ends in compression

fs ¼ k Δx ¼ kðx2  x1 Þ; for x2 > x1 (3.2.3)

In case (c), the spring force of the compressed spring is of the form

fs ¼ k Δx ¼ k ðx1  x2 Þ; for x1 > x2 (3.2.4)

As observed from Figure 3.3.2, in the expression of the relative displacement for a spring
that is either tensioned or compressed, the first displacement is the one that has the same
direction as the spring force at the point of application. This is because the applied forces and
corresponding deformation must be consistent in direction. Hence, Δx ¼ x2  x1 for a
tensioned spring as shown in Figure 3.2.2(b); Δx ¼ x1  x2 for a compressed spring, as
shown in Figure 3.2.2(c). This observation is useful in modeling of mechanical systems.
Note that the forces applied at a massless spring must be balanced at any time. In other
words, the forces at the two ends of a spring are always the same in magnitude and opposite
in direction, as shown in Figures 3.2.2(b) and 3.2.2(c).

Effective Spring Coefficients of Elastic Bodies

Translational springs in Figure 3.2.2 are lumped-spring models that are deduced from elastic
bodies. Indeed, spring elements in engineering applications are made of flexible bodies. Such
a flexible body can be described by the theory of the strength of materials, and a spring
coefficient so determined is called the effective spring coefficient. In this process, it is often
assumed that the inertia of the flexible body is negligible.
As an example, consider a massless cantilever beam subject to a load F at its tip; see Figure
3.2.3(a). By beam theory, the beam tip deflection y is expressed by

L3
y¼ F
3EI
 3.2 Translational Systems 123

Figure 3.2.3 Effective spring coefficient: (a) massless cantilever beam; (b) beam treated as a spring

where EI and L are the bending stiffness and length of the beam, respectively. The previous
equation can be written as

F ¼ keff y

with the effective spring coefficient keff given by

3EI
keff ¼
L3

Thus, the massless cantilever beam can be treated as a spring, as shown in Figure 3.2.3(b).
Table 3.2.1 lists the effective spring coefficients for several elastic continua commonly seen
in engineering applications.

Viscous Dampers
A viscous damper or dashpot, which is simply called a damper, is a device generating a force
(damping force) that is proportional to the difference Δv in the velocity at the two ends of the
device:

Damping force ¼ c  Δv (3.2.5)

where c is a damping coefficient, in N∙s/m. The Δv is also called relative velocity. Figure
3.2.4 shows three cases of dampers in deformation: (a) a damper with a fixed end; (b) a
damper of two movable ends in tension; and (c) a damper of two movable ends in compres-
sion. In case (a), the damping force fd is related to the velocity at the right end of the
damper by

fd ¼ cΔv ¼ cx_ (3.2.6)

in which, Δv ¼ x_  0 ¼ x.
_ In case (b), the damping force of the tensioned damper is
given by

fs ¼ c Δv ¼ c ðx_ 2  x_ 1 Þ; for x_ 2 > x_ 1 (3.2.7)

124 3 Mechanical Systems

Table 3.2.1 Effective spring coefficients of certain continua

Cantilever beam
3EI
F F ¼ keff y; keff ¼
L3
EI, L
EI ¼ bending stiffness
y

Simply supported beam

48EI
F F ¼ keff y; keff ¼
L3
EI
EI ¼ bending stiffness
y

L/2 L/2

Fixed-end beam
192EI
F F ¼ keff y; keff ¼
L3
EI
EI ¼ bending stiffness
y

L/2 L/2

Rod in longitudinal deformation

EA
F ¼ keff u; keff ¼
L
EA ¼ longitudinal rigidity

Taut string
4T
F ¼ keff y; keff ¼
L
T ¼ tension

Rod in torsion
GJ
GJ, L τ ¼ keff θ; keff ¼
L
θ
GJ ¼ torsional rigidity
τ
 3.2 Translational Systems 125

Figure 3.2.4 Dampers in deformation: (a) damper with a fixed end; (b) damper of two movable ends in
tension; (c) damper of two movable ends in compression

In case (c), the damping force of the compressed damper is of the form

fd ¼ c Δv ¼ c ðx_ 1  x_ 2 Þ; for x_ 1 > x_ 2 (3.2.8)

Similar to springs in deformation, the first velocity in the expression of Δv is the one which
has the same direction as the damping force. Thus, Δv ¼ x_ 2  x_ 1 for a tensioned damper, as
shown in Figure 3.2.4(b); Δv ¼ x_ 1  x_ 2 for a compressed damper, as shown in Figure 3.2.4(c).
Also, with the massless assumption for dampers, the forces applied at a damper are balanced
at any time. Consequently, the damping forces at the two ends of a damper are always the
same in magnitude and opposite in direction.

Friction Models
Friction forces are induced by the relative velocity between two moving objects that are in
contact during motion. There are two commonly used models of friction: viscous friction
(fluid friction) and dry friction (Coulomb damping). In viscous friction, as shown in Figure
3.2.5, the friction force is of the form

fd ¼ B  Δv (3.2.9)

where B is a viscous friction coefficient, and Δv is the relative velocity between the objects m1
and m2. Note that Eq. (3.2.9) is the same in format as Eq. (3.2.5) for dampers. Figure 3.2.5(a)
shows a schematic of viscous friction between two moving objects (m1 and m2).
Unlike the damping forces of a damper, the viscous friction forces are the action and
reaction forces that are separately applied to the two objects. Figures 3.2.5(b) and 3.2.5(c)
show two cases of friction forces in free-body diagrams. In each case, the relative velocity is
determined by the relative motion on the two surfaces of the viscous friction layer, which is
denoted by symbol . With the concept of relative velocity presented in Eqs. (3.2.6) to
(3.2.8), Δv is the difference between the velocity that is in the same direction as the friction
126 3 Mechanical Systems

Figure 3.2.5 Viscous friction: (a) schematic; (b) free-body diagrams for v1 > v2 , Δv ¼ v1  v2 ; (c)) free-
body diagrams for v2 > v1 , Δv ¼ v2  v1

force fd and the velocity that is in the opposite direction of the friction force. Here, the
velocity and friction force in consideration must be on the same surface of the viscous friction
layer. Thus, for v1 > v2 , Δv ¼ v1  v2 , as shown in Figure 3.2.5(b); for v2 > v1 , Δv ¼ v2  v1 ,
as shown in Figure 3.2.5(c).
Figure 3.2.6 (a) shows a schematic of dry friction between two moving objects (m1 and m2 ).
The friction force is of the form

μD N; for Δv < 0

fμ ¼ (3.2.10)
μD N; for Δv > 0

where μD is a coefficient of kinematic friction, N is a normal force, and Δv is the relative

velocity between the two objects. For simplicity, Eq. (3.2.10) can written as

fμ ¼ μD N sgnðΔvÞ (3.2.11)

where sgnðaÞ is a sign function, with sgnðaÞ ¼ 1 for a > 0, sgnðaÞ ¼ 1 for a < 0 and
sgnðaÞ ¼ 0 for a = 0.
Figures 3.2.6(b) and 3.2.6(c) show two cases of a dry-friction force fμ and a normal force N
in free-body diagrams. In each case, the relative velocity is determined by the relative motion
on the two surfaces of the dry-friction region, which is denoted by symbol . With the
concept of relative velocity presented in Eqs. (3.2.6) to (3.2.8), Δv is the difference between
the velocity that is in the same direction as the friction force fμ and the velocity that is in the
opposite direction of the friction force. Hence, the velocity and friction force in consideration
must be on the same surface of the dry-friction region. Thus, for v1 > v2 , Δv ¼ v1  v2 , as
shown in Figure 3.2.6(b); and for v2 > v1 , Δv ¼ v2  v1 , as shown in Figure 3.2.6(c).
 3.2 Translational Systems 127

v2 v2
m2 m2
fμ fμ
Dry v2 N N
friction N N
(μ D ) m2 fμ v2 fμ v2

m1 fμ v1 fμ v1
N N
v1 N N
fμ fμ
(a)
m1 m1

v1 v1
(b) (c)

Figure 3.2.6 Dry friction: (a) schematic; (b) free-body diagrams for v1 > v2 , Δv ¼ v1  v2 ; (c) free-body
diagrams for v2 > v1 , Δv ¼ v2  v1

As seen from Eq. (3.2.10), dry-friction forces are nonlinear functions of the relative motion
between two objects. Different from viscous friction, the magnitude of a dry-friction force fμ
is not related to the relative velocity Δv, but it depends on the normal force N. Also, as
depicted in Figures 3.2.6(b) and 3.2.6(c), dry-friction forces are the action and reaction forces
that are separately applied to the two objects.
In summary of this subsection, a translational system consists of the above-described
lumped masses, springs, dampers, and friction models. Because of this, such a system is often
called a spring–mass–damper system. There are other models of spring and damping
elements, including nonlinear springs and air damping, some of which shall be introduced
later on.

3.2.2 Free-Body Diagrams and Auxiliary Plots

In Sections 3.1 and 3.2.1, the first two keys in system modeling, namely the fundamental
principles and basic elements, have been presented. In this section, we consider the last
(third) key – ways of analysis. In a way of analysis, the fundamental principles are applied to
basic elements, such that the governing equations of a mechanical system in consideration
are obtained. There are two typical ways of analysis: free-body diagrams and Lagrange’s
equations. The former is part of Newtonian mechanics and the latter belongs to Lagrangian
mechanics. In this section, free-body diagrams are considered. Lagrange’s equations shall be
introduced in Section 3.5.
Free-body diagrams are studied in a typical undergraduate dynamics course. A free-body
diagram is the diagram of a mass element showing all internal and external forces applied to
the element and marking the coordinates or displacements of the element. Here, internal
128 3 Mechanical Systems

k x (t )
fs

x (t ) x (t ) x (t )
k
F(t ) fs F (t ) fs F (t )
m m m
μ fd fd
c
fμ fμ
N
mg
(a) (b) (c)

Figure 3.2.7 Free-body diagrams of a spring-mass-damper system (see text for details)

forces refer to those forces applied by spring and damping elements (including friction
forces) while external forces, including gravity, are those that are not internal forces. In
generating a free-body diagram, no connection between the mass element in consideration
and any other element is allowed. In other words, for a system with N particles, N free-body
diagrams should be drawn.
As an example, consider a spring–mass–damper system in Figure 3.2.7(a), which is
constrained by a spring (k), a damper (c), and dry friction ( μ), subject to an external force
(F ), and under the gravity. The two-dimensional free-body diagram of the mass m is shown
Figure 3.2.7(b), where fs , fd , and fμ are the internal forces of the spring, damper, and dry
friction, mg is the gravity, and N is the normal force applied by the supporting horizontal
surface. If only the motion in the horizontal direction is of interest, the free-body diagram
can be simplified to show quantities only in the x direction; see Figure 3.2.7(c).
For convenience in system modeling, plots of spring and damping elements can be drawn,
which are called auxiliary plots. The plots in Figures 3.2.2 and 3.2.4 are examples of auxiliary
plots. Like a free-body diagram, the auxiliary plot of a spring or damping element concerns
the forces applied to the element and it does not show the connection between the element
and any other elements. Auxiliary plots are usually needed for spring/damping elements with
two movable ends.
Auxiliary plots are especially useful in modeling of a system with multiple mass elements.
For instance, consider the spring–mass–damper system in Figure 3.2.8(a), where two masses
ðm1 ; m2 Þ are connected to two springs ðk1 ; k2 Þ and one damper (c). Figure 3.2.8(b) shows two
free-body diagrams about the masses, and one auxiliary plot about the spring of coefficient
k2 , where, by Newton’s third law on action and reaction, the directions of the internal forces
are properly assigned. Note that the spring is tensioned with the assumed direction of the
spring force fs2 , indicating that the relative displacement Δx ¼ x2  x1 by Eq. (3.2.3).
Since the displacements x1 and x2 are unknown, the spring is not necessarily tensioned, and
it can also be compressed. Figure 3.2.8(c) shows an auxiliary plot of a compressed spring k2 .
 3.2 Translational Systems 129

Figure 3.2.8 Free-body diagrams and auxiliary plots for a spring–mass–damper system (see text for
details)

In this case, the spring force fs2 is in the opposite direction to that in Figure 3.2.7(b) and the
relative displacement becomes Δx ¼ x1  x2 by Eq. (3.2.4). In system modeling, either a
tensioned spring or a compressed spring can be assumed, and the governing equations
eventually obtained should be the same. This is because the change of the direction of a
spring force is matched by the change of sign of the corresponding relative displacement Δx.
In summary, with a system of N particles, N free-body diagrams must be drawn and some
auxiliary plots of spring and damping elements may be generated. An auxiliary plot is
necessary only if the particular spring or damping element has two movable ends with
unknown displacements.
130 3 Mechanical Systems

3.2.3 Derivation of Equations of Motion

With the three keys of system modeling (fundamental principles, basic elements, and free-
body diagrams) presented in Sections 3.1, 3.2.1, and 3.2.3, the governing equations for a
translational system are readily established. This process of system modeling takes the
following four steps:

(1) Draw free-body diagrams, one for each mass element, and create auxiliary plots for some
spring/damping elements.
(2) Apply Newton’s second law to each free-body diagram, yielding the differential equa-
tions about the displacements of the mass elements.
(3) Write down the expressions of internal forces (of the spring and damping elements) and
use auxiliary plots if necessary.
(4) Obtain the governing equations of motion of the system through substitution of the
expressions of internal forces given in Step 3 into the equations obtained in Step 2.

As a convention, the equations of motion so obtained in the above-described process list

the terms of unknown displacements and their derivatives on the left-hand sides, and the
terms about external forces on the right-hand sides.
In modeling of a mechanical system, a spring or damping element can be assumed either in
tension or in compression. The final equations of motion of the system are always the same,
provided that the corresponding internal forces and relative motion are treated consistently.

Example 3.2.1
For the spring–mass–damper system in Figure 3.2.8(a), derive the equations of motion.
Solution
The free-body diagrams of the masses and an auxiliary plot of the spring k2 are drawn in Figure
3.2.8(b), where the spring is assumed to be tensioned. By Newton’s second law and with the free-
body diagrams, the governing equations of the masses are obtained as follows

For m1 m1 x€1 ¼ f1 þ fs2  fs1  fd (a)

For m2 m2 x€2 ¼ f2  fs2 (b)

According to Section 3.2.1 and the auxiliary plot in Figure 3.2.8(b), the internal forces are
expressed by

f1 ¼ k1 x1
fd ¼ cx_ 1 (c)
fs2 ¼ k2 ðx2  x1 Þ

where Eq. (3.2.3) has been used to obtain the expression of fs2 . Substituting Eq. (c) into Eqs. (a)
and (b) yields
 3.2 Translational Systems 131

m1 x€1 ¼ f1 þ k2 ðx2  x1 Þ  k1 x  cx_ 1

(d)
m2 x€2 ¼ f2  k2 ðx2  x1 Þ

Finally, rearranging the previous equations gives the equations of motion of the system
m1 x€1 þ cx_ 1 þ ðk1 þ k2 Þx1  k2 x2 ¼ f1
(e)
m2 x€2 þ k2 x2  k2 x2 ¼ f2

If the spring k2 is assumed to be compressed, the corresponding free-body diagrams and

auxiliary plot are shown in Figure 3.2.8(c). The differential equations by Newton’s second law
become

For m1 m1 x€1 ¼ f1  fs2  fs1  fd (f)

For m2 m2 x€2 ¼ f2 þ fs2 (g)

in which the signs of the fs2 -terms have changed. On the other hand, the force of the spring k2 by
Eq. (3.2.4) becomes fs2 ¼ k2 ðx1  x2 Þ, which differs from the spring force expression in Eq. (c) by a
sign. The other internal forces remain the same. This indicates that the same equations (e) of
motion can be obtained via substitution and manipulation.

Example 3.2.2
In Figure 3.2.9, a mass is constrained by a spring and a damper that are serially connected. Let the
displacement at the point of the spring–damper connection be y(t). Derive the equations of motion
for the translational system.
Solution
Figure 3.2.10 shows the free-body diagram of the mass and the auxiliary plots of the damping and
spring elements. Note that the spring and damper are in tension. The equations about the three
basic elements are as follows:

Figure 3.2.9 A mass constrained by serially connected spring and damper

132 3 Mechanical Systems

Figure 3.2.10 Free-body diagram and auxiliary plots for the system in Example 3.2.2

Mass: x ¼ F  fs
m€
Damper: fd ¼ cy_
Spring: fs ¼ kðx  yÞ

where Eq. (3.2.3) has been used for the tensioned spring. Because the spring and damper are
connected, fd ¼ fs . It follows that the system is governed by the coupled differential equations

m€ x þ kx  ky ¼ F
cy_ þ ky  kx ¼ 0

Note that the displacements x and y are independent of each other even though no mass element
related to y exists. The coupled differential equations can be solved by the Laplace transform
(Section 2.4) or a transfer function formulation (Section 3.2.4).

Example 3.2.3
In Figure 3.2.11, a mass M slides on a slope of angle α, and is constrained by a spring k. There exists dry
friction ðμD Þ between the slope and mass and gravity is naturally considered. Let x be the
displacement of the mass on the slope. Derive the equation of motion of the mass.

k
x (t )
M
g
F (t )
μD

Figure 3.2.11 A mass sliding on a slope

 3.2 Translational Systems 133

Solution
The free-body diagram of the mass is shown in Figure 3.2.12, where axis y is normal to the
slope surface. By Newton’s second law, the motion of the mass in the x direction is
governed by

x ¼ F þ Mg sin α  fs  fμ
m€

y (t )

fs
x (t )
fμ M

F (t )

N Mg

Figure 3.2.12 Free-body diagram of the mass in Example 3.2.3

_ is given by
The spring force is fs ¼ kx. The dry-friction force, by Eq. (3.2.11) with Δv ¼ x,

_
fμ ¼ μD N sgnðxÞ

Because the mass does not have motion in the y direction, N ¼ Mg cos α. Therefore, the equation
of motion of the mass is

_ ¼ F þ Mg sin α
x þ kx þ μD Mg cos α sgnðxÞ
m€

which is a nonlinear differential equation.

Example 3.2.4
A translational model of an automotive suspension system is shown in Figure 3.2.13, which is
known as quarter-car suspension model because only one wheel is considered. In the figure, spring
k1 models the stiffness of the tire; mass m1 represents the inertia of the tire and axle, spring k2 and
damper c describe the suspension system, mass m2 describes the inertia of the supported vehicle
components, including the chassis structure, engine, and driver and passengers; and y0 ðtÞ
134 3 Mechanical Systems

Figure 3.2.13 A quarter-car suspension model

prescribes the surface profile of the road. By considering gravity, derive the equations of motion of
the car suspension system.
Solution
The free-body diagrams and auxiliary plots of the system are drawn in Figure 3.2.14, where all the
spring and damping elements are in tension. By the free-body diagrams, the differential equations
about the displacements y1 ðtÞ and y2 ðtÞ are given by

m1 y€1 ¼ fs2 þ fd  m1 g  fs1

(a)
m2 y€2 ¼ m1 g  fs2  fd

The internal forces, by the auxiliary plots, are determined as

fs1 ¼ k1 ðy1  y0 Þ
fs2 ¼ k2 ðy2  y1 Þ (b)
fd ¼ cðy_ 2  y_ 1 Þ

By substituting Eq. (b) into Eq. (a), the governing equations of motion of the vehicle system are
obtained as follows:

m1 y€1 þ cy_ 1 þ ðk1 þ k2 Þy1  cy_ 2  k2 y2 ¼ k1 y0  m1 g

(c)
m2 y€2 þ cy_ 2 þ k2 y2  cy_ 1  k2 y1 ¼ m2 g
 3.2 Translational Systems 135

Figure 3.2.14 Free-body diagrams and auxiliary plots of the quarter-car suspension model

where the prescribed road surface profile y0 serves as a disturbance to the vehicle system.

Example 3.2.5
In Figure 3.2.15, a cart of mass m1 carries a lumped mass m2 and it is subject to an external force f.
The cart and the mass are connected by a spring (k2) and there is viscous friction (B) between the
two bodies. Develop a mathematical model for the cart–mass system. Also, the cart is constrained
by another spring (k1)

Figure 3.2.15 A cart–mass system in Example 3.2.5

Solution
Figure 3.2.16 shows the free-body diagrams and the auxiliary plots of the system. According to the
free-body diagrams, the differential equations of the bodies are
136 3 Mechanical Systems

m1 x€1 ¼ f  fs1 þ fs2  fd

m2 x€2 ¼ fd  fs2
m1 x€1 ¼ f  fs1 þ fs2  fd
m2 x€2 ¼ fd  fs2

By the auxiliary plots, the internal forces are determined as

fs1 ¼ k1 x1 ; fs2 ¼ k2 ðx2  x1 Þ; fd ¼ Bðx_ 1  x_ 2 Þ
It follows that
m1 x€1 ¼ f  k1 x1 þ k2 ðx2  x1 Þ  Bðx_ 1  x_ 2 Þ
m2 x€2 ¼ Bðx_ 1  x_ 2 Þ  k2 ðx2  x1 Þ

which, after rearrangement, leads to the mathematical model of the cart–mass system

m1 x€1 þ Bx_ 1 þ ðk1 þ k2 Þx1  Bx_ 2  k2 x2 ¼ f

m2 x€2 þ Bx_ 2 þ k2 x2  Bx_ 1  k2 x1 ¼ 0

Figure 3.2.16 Free-body diagrams and auxiliary plots for the system in Example 3.2.5

In the previous examples, constant masses are considered. For mechanical systems with time-
varying inertia, such as rockets with burning fuel, Eq. (3.1.2) is not directly applicable. A system
with time-varying parameters is known as a time-variant system. In this case, one may consider
the principle of impulse and momentum for a system of particles, as presented in Section 3.1.5.
Indeed, the principle of impulse and momentum is the result of the temporal integration of
Newton’s second law and it serves as a fundamental principle in system modeling. See the next
example.

Example 3.2.6
In Figure 3.2.17(a), a rocket is moving upward against gravity. At time t, the rocket has mass mR ðtÞ
and velocity vðtÞ. Let the effective exhaust velocity of burning fuel be ue ðtÞ, which is measured from
the rocket frame. Derive the equation of motion for the rocket.
 3.2 Translational Systems 137

Solution
Assume that a small mass dm is expelled from the rocket during the time interval dt. Thus, at time
t þ dt, we have a two-particle system: the rocket with mass mR  dm and velocity v þ dv, and the
expelled mass dm with velocity v  ue ; see Figure 3.2.17(b). Let fdrag be a drag force acting on the
rocket body (not shown in Figure 3.2.17). The impulse of the gravity and drag force and the linear
momentums at t and t þ dt, by Eq. (3.1.22), are

I 12 ¼ ðmR g þ fdrag Þ dt
(a)
L1 ¼ mR v; L2 ¼ ðmR  dmÞðv þ dvÞ þ dm ðv  ue Þ

According to the principle of impulse and momentum, I 12 ¼ L2  L1 , we have

ðmR g þ fdrag Þdt ¼ ðmR  dmÞðv þ dvÞ þ dm ðv  ue Þ  mR v (b)

which, after neglecting the higher-order term dmdv, is reduced to

ðmR g þ fdrag Þdt ¼ mR dv  ue dm (c)

Because the mass of the rocket during the interval dt is decreased, dmR ¼ dm, Eq. (c) is reduced to

ðmR g þ fdrag Þdt ¼ mR dv þ ue dmR (d)

It follows from Eq. (d) that the equation of motion of the rocket is

dv dmR
mR ¼ ue  mR g  fdrag (e)
dt dt
dmR
where the term ue is called rocket thrust. This equation is known as the Tsiolkovsky rocket
dt
equation or the ideal rocket equation.

Figure 3.2.17 The inertia and velocity of a rocket at times (a) t and (b) t þ dt
138 3 Mechanical Systems

Example 3.2.7
In Figure 3.2.18(a), a massless simply supported beam is carrying a spring–mass system at its
midpoint. Assume that mass m1 has the same deflection as the beam at the contact point. This
system can be a model of automobile chassis system. With consideration of gravity, develop a
mathematical model of the system.

Figure 3.2.18 Simply supporting beam carrying a spring–mass system: (a) schematic; (b) model with
the beam treated as a spring

Solution
The massless beam can be treated as a spring of effective coefficient, which by Table 3.2.1 is given
by keff ¼ 48EI=L3 . Thus, the combined beam–mass system can be approximated by the one in
Figure 3.2.18(b), where y1 and y2 are the displacements of the masses. Following the steps in
Example 3.2.4 or Example 3.2.5, it can be shown that the governing equations of the combined
system are

m1 y€1 þ ðkeff þ kÞy1  ky2 ¼ m1 g

m2 y€2 þ ky2  ky1 ¼ m2 g

3.2.4 Transfer function Formulation

As mentioned in Section 1.3, there are four commonly used forms of model representation
for modeling, simulation and analysis of dynamic systems: (i) transfer function formulation,
(ii) state-space representation, (iii) block diagram representation in the s-domain, and (iv)
block diagram representation in the time domain. In this section, a transfer function
 3.2 Translational Systems 139

formulation for mechanical systems is introduced. Block diagram and state-space representa-
tions are presented in Sections 3.6 and 3.7. These forms of model representation are convenient
for computation with MATLAB and Mathematica, as is shown in Sections 3.8 and 3.9.
A transfer function formulation is a method for modeling and analysis of time-invariant
dynamic systems whose parameters do not change with time. As discussed in Sections 1.3
and 2.7, the transfer function of a time-invariant system is defined as the ratio of the system
output to the system input in the Laplace transform domain with zero initial conditions:


L½output


GðsÞ ¼ (3.2.12)
L½input
zero initial conditions

where L is the Laplace transform operator, and zero initial conditions mean all the initial
values of the input and output and their time derivatives are set to zero. Thus, a transfer
function represents an input–output relation for the system. With a transfer function
formulation, the characteristics of a dynamic system, including the system response and
stability, can be evaluated, and feedback control laws can be designed and implemented.
These topics will be explored in Chapters 7 and 8.
The process of the establishment of transfer functions is called a transfer function formu-
lation. In general, a transfer function can be expressed as

BðsÞ bm sm þ bm1 sm1 þ . . . þ b1 s þ b0

GðsÞ ¼ ¼ (3.2.13)
AðsÞ an sn þ an1 sn1 þ . . . þ a1 s þ a0

where A(s) and B(s) are polynomials of the Laplace parameter s. The order of the transfer function
is the order of the denominator polynomial A(s). Thus, G(s) in Eq. (3.2.2) is an n-th-order transfer
function. A system with an n-th-order transfer function is called an n-th-order system.
For a mechanical system, its input can be an external force or a prescribed displacement, its
output can be a displacement or velocity or an internal force. The transfer function for a given
physical system is not unique, depending on the specification of input and output. Also, a system
may have multiple inputs and multiple outputs. For such a system, multiple transfer functions
can be established, with a transfer function for one pair of input and output.
The derivation of a transfer function for a mechanical system generally takes the following
steps:

(1) Define a transfer function by selecting an input and an output.

(2) Derive the time-domain governing equations of motion for the system.
(3) Take the Laplace transform of the time-domain equations with zero initial conditions.
(4) With the Laplace transform equations, determine the ratio of a relevant unknown
function to a given input function, which may or may not be the transfer function.
(5) By the result obtained in Step 4 and the selected input and output in Step 1, obtain the
transfer function of the system.

Of course, Step 5 is not necessary if the ratio obtained in Step 4 is the transfer function.
140 3 Mechanical Systems

Example 3.2.8
For the spring–mass–damper system in Figure 3.2.19, derive the transfer function with the
_ of the mass as the output.
external force f ðtÞ as the input and the velocity xðtÞ

Figure 3.2.19 A spring–mass–damper system

Solution
The transfer function, by Eq. (3.2.12), is of the form


_


L½x sX ðsÞ
GðsÞ ¼ ¼ (a)
L½ f 
zero initial conditions FðsÞ

It is easy to show that the equation of motion of the system is

_ þ kxðtÞ ¼ f ðtÞ
x ðtÞ þ cxðtÞ
m€

Performing Laplace transform of the previous differential equation with zero initial conditions gives
 2 
ms þ cs þ k X ðsÞ ¼ FðsÞ

from which
X ðsÞ 1
¼ 2 (b)
FðsÞ ms þ cs þ k

The ratio X ðsÞ=FðsÞ is not the transfer function as defined in Eq. (a). However, it is useful in the
determination of the transfer function as shown below:

sX ðsÞ s
GðsÞ ¼ ¼ 2 (c)
FðsÞ ms þ cs þ k

This is a transfer function of order two. Thus, the system is a second-order system.

Example 3.2.9
For the spring–mass–damper system shown in Figure 3.2.9, derive the transfer function with
external force F as the input and displacement x as the output. Also, derive another transfer
function with F as the input and the spring force as the output.
 3.2 Translational Systems 141

Solution
With the defined input and output, the first transfer function is GðsÞ ¼ X ðsÞ=FðsÞ. Taking Laplace
transform of the system governing equations in Example 3.2.2, with zero initial conditions, gives
the s-domain algebraic equations about X ðsÞ and YðsÞ

ðms2 þ kÞX ðsÞ  kYðsÞ ¼ FðsÞ

(a)
ðcs þ kÞYðsÞ  kX ðsÞ ¼ 0

Note that X ðsÞ is in the definition of the transfer function. By the second line in Eq. (a)
k
YðsÞ ¼ X ðsÞ (b)
cs þ k

Subsitituting Eq. (b) into the first line of Eq. (a) gives

1
X ðsÞ ¼ FðsÞ
k2
ðms2 þ kÞ 
cs þ k

which yields the transfer function of the system as follows:

X ðsÞ cs þ k
GðsÞ ¼ ¼
FðsÞ ðms2 þ kÞðcs þ kÞ  k2
(c)
cs þ k
¼
sðmcs2 þ mks þ kcÞ
Equation (c) represents a third-order system.
If the output is specified as the spring force fs ¼ kðx  yÞ, with the input being the same, the
second transfer function is defined by
 
Fs ðsÞ k X ðsÞ  YðsÞ
G1 ðsÞ ¼ ¼ (d)
FðsÞ FðsÞ

In this case, both X ðsÞ and YðsÞ must be obtained before the transfer function can be determined.
By Eqs. (b) and (c),

cs þ k k
X ðsÞ ¼ FðsÞ; YðsÞ ¼ X ðsÞ (e)
sðmcs2 þ mks þ kcÞ cs þ k

It follows that the transfer function

 
X ðsÞ YðsÞ k X ðsÞ
G1 ðsÞ ¼ k  ¼ 1
FðsÞ FðsÞ cs þ k FðsÞ
 (f)
cs þ k k ck
¼k 1 ¼
sðmcs þ mks þ kcÞ
2 cs þ k mcs þ mks þ kc
2

Unlike GðsÞ, G1 ðsÞ is a second-order transfer function.

142 3 Mechanical Systems

Example 3.2.9 shows that a dynamic system can have different transfer functions with
different orders, depending on the specification of input and output.

Example 3.2.10
For the quarter-car suspension model in Example 3.2.4, derive the transfer function with the
prescribed road-surface profile y0 as the input and the displacement y2 as the output. Also, identify
the order of the system.
Solution
The transfer function is expressed by Y2 ðsÞ=Y0 ðsÞ. Taking Laplace transform of Eq. (c) of motion
in Example 3.2.4 gives

ðm1 s2 þ cs þ k1 þ k2 ÞY1 ðsÞ  ðcs þ k2 ÞY2 ðsÞ ¼ k1 Y0 ðsÞ

(a)
ðm2 s2 þ cs þ k2 ÞY2 ðsÞ  ðcs þ k2 ÞY1 ðsÞ ¼ 0

where the gravitational forces have been eliminated because they are not related to the defined
transfer function. (A transfer function can only take one input and one output.) By substitution
m2 s2 þ cs þ k2
Y1 ðsÞ ¼ Y2 ðsÞ
cs þ k2
m2 s2 þ cs þ k2
ðm1 s2 þ cs þ k1 þ k2 Þ Y2 ðsÞ  ðcs þ k2 ÞY2 ðsÞ ¼ k1 Y0 ðsÞ
cs þ k2

which eventually yields the transfer function as follows

1
Y2 ðsÞ m2 s2 þ cs þ k2
¼ k1 ðm1 s2 þ cs þ k1 þ k2 Þ  ðcs þ k2 Þ
Y0 ðsÞ cs þ k2
k1 ðcs þ k2 Þ
¼
ðm1 s2 þ k1 þ cs þ k2 Þðm2 s2 þ cs þ k2 Þ  ðcs þ k2 Þ2
k1 ðcs þ k2 Þ
¼
ðm1 s2 þ k1 Þðm2 s2þ cs þ k2 Þ þ m2 s2 ðcs þ k2 Þ

Because the transfer function is of order four, the quarter-car suspension model is a fourth-
order dynamic system.

Example 3.2.11
For the spring–mass–damper system in Figure 3.2.8(a), determine the transfer functions
X2 ðsÞ=F1 ðsÞ and X2 ðsÞ=F2 ðsÞ.
Solution
There are two methods for determining the transfer functions:

Method 1. First set F2 ðsÞ ¼ 0 and obtain X2 ðsÞ=F1 ðsÞ; and then set F1 ðsÞ ¼ 0 and obtain
X2 ðsÞ=F2 ðsÞ. Each time, a pair of input and output is considered.
 3.2 Translational Systems 143

Method 2. First obtain the complete solutions X2 ðsÞ in terms of F1 ðsÞ and F2 ðsÞ, and then
determine the transfer functions by properly selecting an input.

For comparison purposes, both the methods are used. To this end, take Laplace transform of Eq.
(e) in Example 3.2.1, to obtain

ðm1 s2 þ cs þ k1 þ k2 ÞX1 ðsÞ  k2 X2 ðsÞ ¼ F1 ðsÞ

(a)
ðm2 s2 þ k2 ÞX2 ðsÞ  k2 X1 ðsÞ ¼ F2 ðsÞ

Method 1. For X2 ðsÞ=F1 ðsÞ, setting F2 ðsÞ ¼ 0 reduces Eq. (a) to

ðm1 s2 þ cs þ k1 þ k2 ÞX1 ðsÞ  k2 X2 ðsÞ ¼ F1 ðsÞ

(b)
ðm2 s2 þ k2 ÞX2 ðsÞ  k2 X1 ðsÞ ¼ 0

By substitution
m2 s2 þ k2
X1 ðsÞ ¼ X2 ðsÞ
k2
m2 s2 þ k2
ðm1 s2 þ cs þ k1 þ k2 Þ X2 ðsÞ  k2 X2 ðsÞ ¼ F1 ðsÞ
k2
which leads to the transfer function
X2 ðsÞ k2
¼
F1 ðsÞ ðm1 s þ cs þ k1 þ k2 Þðm2 s2 þ k2 Þ  k22
2
(c)
k2
¼
ðm1 s2 þ cs þ k1 Þðm2 s2 þ k2 Þ þ m2 k2 s2

Likewise, setting F2 ðsÞ ¼ 0 reduces Eq. (a) to

ðm1 s2 þ cs þ k1 þ k2 ÞX1 ðsÞ  k2 X2 ðsÞ ¼ 0

(d)
ðm2 s2 þ k2 ÞX2 ðsÞ  k2 X1 ðsÞ ¼ F2 ðsÞ

By substitution
k2
X1 ðsÞ ¼ X2 ðsÞ
m1 þ cs þ k1 þ k2
s2
k2
ðm2 s2 þ k2 ÞX2 ðsÞ  k2 X2 ðsÞ ¼ F2 ðsÞ
m1 s þ cs þ k1 þ k2
2

which leads to the transfer function

X2 ðsÞ m1 s2 þ cs þ k1 þ k2
¼ (e)
F2 ðsÞ ðm1 s2 þ cs þ k1 Þðm2 s2 þ k2 Þ þ m2 k2 s2

Method 2. The coupled algebraic equations in Eq. (a) are written in the matrix form
 
αðsÞ k2 X1 ðsÞ F1 ðsÞ
¼ (f)
k2 βðsÞ X2 ðsÞ F2 ðsÞ
144 3 Mechanical Systems

where
   
αðsÞ ¼ m1 s2 þ cs þ k1 þ k2 ; βðsÞ ¼ m2 s2 þ k2

Application of Cramer’s rule to Eq. (f) yields

1

αðsÞ

F1 ðsÞ

X2 ðsÞ ¼


ΔðsÞ
k2 F2 ðsÞ

1
¼ ½k2 F1 ðsÞ þ αðsÞF2 ðsÞ (g)
ΔðsÞ

where ΔðsÞ is the determinant of the matrix in Eq. (f ) and it is given by

ΔðsÞ ¼ αðsÞβðsÞ  k22

(h)
¼ ðm1 s2 þ cs þ k1 Þðm2 s2 þ k2 Þ þ m2 k2 s2
k2
For F2 ðsÞ ¼ 0, Eq. (g) yields X2 ðsÞ=F1 ðsÞ ¼ , which by Eq. (h) gives
ΔðsÞ
X2 ðsÞ k2
¼ (i)
F1 ðsÞ ðm1 s þ cs þ k1 Þðm2 s2 þ k2 Þ þ m2 k2 s2
2

αðsÞ
For F1 ðsÞ ¼ 0, Eq. (g) yields X2 ðsÞ=F2 ðsÞ ¼ , which gives
ΔðsÞ
X2 ðsÞ m1 s2 þ cs þ k1 þ k2
¼ (j)
F2 ðsÞ ðm1 s2 þ cs þ k1 Þðm2 s2 þ k2 Þ þ m2 k2 s2

Obviously, the transfer functions obtained by the two methods are the same.

As mentioned previously, a transfer function formulation is only valid for linear time-
invariant systems with linear governing equations having constant coefficients. This is
because the Laplace transform operator is linear and because it is not applicable to products
of time functions. Hence, no transfer functions can be derived for nonlinear systems (like the
system subject to dry friction in Example 3.2.3) and linear time-varying systems (like the
rocket in Example 3.2.6). For linear time-varying systems and nonlinear systems, a state-
space representation can be used for simulation and analysis; see Section 3.7.

3.3 Rotational Systems

A rotational system is a mechanical system with its components constrained to rotate about an
axis. Examples include rotating shafts, spinning disks, slewing arms, pulley systems, and geared
systems. In this section, rigid bodies rotating about a fixed axis are considered. General motion
of rigid bodies, involving both translation and rotation, is discussed in Section 3.4.
 3.3 Rotational Systems 145

3.3.1 Three Keys in the Derivation of the Equations of Motion

The derivation of the equations of motion for rotational systems, as in modeling of transla-
tional systems, have three keys: fundamental principles, basic elements, and a way of analysis
(free-body diagrams and auxiliary plots).

Double-Headed Arrow for Rotation

The description of the rotation of a body involves an angle of rotation and torques/moments
that cause the rotation. In Figure 3.3.1(a), a rigid body, which is subject to a torque τ and a
force F, rotates about a fixed axis O–O with an angle θ of rotation. For convenience in
modeling and analysis, the rotation (angular displacement, velocity, and acceleration) and
applied forces are represented by double-headed arrows (double-headed vectors); see Figure
3.3.1(b), where ω ¼ θ,_ α ¼ θ,
€ and MO is the moment of force F. The direction of a double-
headed arrow is determined by the right-hand rule, and the summation of the double-headed
arrows is treated as vector summation. Double-headed arrows are useful in dealing with the
interactions of multiple rotating bodies. In this text, the conventional notation for rotation
and double-headed arrows are in mixed use whenever it is convenient.

Fundamental Principles
Rigid bodies in rotation about a fixed axis are commonly seen in two cases: (i) a three-
dimensional rigid body rotating about an axis, as in Figure 3.3.1; and (ii) a slab rotating

τ
O O

F
(a)

ω α
τ MO

(b)

Figure 3.3.1 A rotating rigid body: (a) schematic; (b) double-headed arrows by the right-hand rule
146 3 Mechanical Systems

θ O
τ

Figure 3.3.2 A slab rotates about hinge O

about a hinge, as shown in Figure 3.3.2, where the rotation axis is perpendicular to the paper,
and goes through the hinge point O. A rotating shaft is an example in case (i) and a spinning
disk is an example in case (ii). In either case, Newton’s second law for particles, Eq. (3.1.2), can
be extended as follows

IO θ€ ¼ MO (3.3.1)
where MO is the resultant of torques and moments of forces about the axis or hinge, and IO is
the mass moment of inertia of the body about the axis or hinge. The mass moment of inertia
of a body is defined as ð
IO ¼ r2 dm (3.3.2)
V

where r is the distance from the axis or hinge to the infinitesimal mass element dm, and V is
the volume of the body.

Parallel-Axis Theorem Assume that a body of mass m has a mass moment of inertia IG with
respect to the axis A that passes through the mass center G of the body. (Refer to Section
3.4.1 for the definition of the mass center.) For the body rotating about axis O that is parallel
to axis A at a distance d (see Figure 3.3.3), its mass moment of inertia with respect to axis O
can be expressed by
IO ¼ IG þ md2 (3.3.3)

As in Sections 3.1.4 and 3.1.5, the spatial and temporal integration of Eq. (3.3.1) yields the
following two principles for a body in rotation.

The Principle of Work and Energy

θð2





1 2
1 2

MO dθ ¼ IO ω
 IO ω
(3.3.4)
2 θ2 2

θ1 θ1
 3.3 Rotational Systems 147

Figure 3.3.3 Parallel axes of a rotating body

_ which is the angular velocity of the body. The left-hand side of Eq. (3.3.4) is the
where ω ¼ θ,
work done by the resultant moment during the rotation of the body from angle θ1 to angle θ2 ,
1
and IO ω2 is the kinetic energy of the body in rotation.
2
The Principle of Impulse and Momentum

ðt2


MO dt ¼ IO ω
t2  IO ω
t1 (3.3.5)
t1

where the left-hand side term is the impulse of the resultant moment in the time interval from
t1 to t2 , and IO ω is the angular momentum of the rotating body. These principles can be used
to develop mathematical models of rigid bodies in rotation.

Basic Elements

Mass Elements A rotational system has three types of basic elements: mass elements, spring
elements, and damping elements. Mass elements are those bodies to which the rotational
version of Newton’s second law, Eq. (3.3.1), can be applied. The bodies in Figures 3.3.1 and
3.3.2 are two examples of mass elements. Other examples of mass elements include rotating
shafts, spinning disks, slewing robot arms, pulleys, and pendulums, which are discussed in
the subsequent sections.

Spring Elements Torsional springs are spring elements for rotational systems. A torsional
spring is a device that produces a resisting torque τs (spring torque) given by
148 3 Mechanical Systems

kT
τs

θ
τs
kT

(a) (b)

Figure 3.3.4 Torsional springs with one end fixed: (a) spring for three-dimensional bodies; (b) spring for
slabs

τs ¼ kT Δθ (3.3.6)

where kT is a torsional spring coefficient, and Δθ is relative rotation, which is the difference of
the angles of twist at the two ends of the spring. Figure 3.3.4 shows two torsional spring
models with one fixed end. The spring in Figure 3.3.4(a) is used with three-dimensional rigid
bodies like the one in Figure 3.3.1(a). The spring in Figure 3.3.4(b) is for slabs like that in
Figure 3.3.2. For either spring, the spring torque is given by

τ s ¼ kT θ (3.3.7)

Torsional springs with two movable ends are usually seen in the rotation of three-
dimensional rigid bodies. Unlike translational springs, no tension and compression can be
identified for torsional springs. Figure 3.3.5 shows two cases of the application of spring
torques. In Figure 3.3.5(a), the double-headed vectors of torques at the two ends of the
spring are pointing away from each other and the relative rotation or twist Δθ ¼ θ2  θ1 . In
Figure 3.3.5(b), the double-headed vectors of torques at the two ends of the spring are
pointing toward each other and the relative rotation Δθ ¼ θ1  θ2 . As observed from the
figure, the first angle in the expression of relative rotation Δθ is the one that has the same
direction as the torque τs at the point of application. This is because the applied torques and
corresponding twist must be consistent in direction. Thus, by Eq. (3.3.6),

τs ¼ kT ðθ2  θ1 Þ (3.3.8a)

for the end torques pointing away from each other as in Figure 3.3.5(a), and by

τs ¼ kT ðθ1  θ2 Þ (3.3.8b)

for the end torques pointing toward each other as Figure 3.3.5(b).
As shown in Table 3.2.1, a massless rod in torsional deformation can be viewed as a
torsional spring with the effective coefficient keff ¼ GJ=L.
 3.3 Rotational Systems 149

τs kT τs

θ1 θ2
(a)

τs kT τs

θ1 θ2
(b)

Figure 3.3.5 Relative rotation of a torsional spring with two movable ends: (a) Δθ ¼ θ2  θ1 ;
(b) Δθ ¼ θ1  θ2

Damping Elements There are several types of damping element models for rotational
systems. A rotational viscous damper is a device with two moving parts that are separated
by a thin film of lubricant; see Figure 3.3.6 (a). Due to viscous friction of the lubricant, the
damper produces a resisting torque τd (damping torque) that is proportional to the relative
angular velocity Δω between the two parts.

τd ¼ bΔω (3.3.9)

where b is a torsional damping coefficient. Shown in Figures 3.3.6(b) and (c) are two cases of
the relative angular velocity, which depends on the direction of the damping torque. Similar
to the relative angular displacement of a torsional spring, the first angular velocity in the
expression of Δω for the rotational viscous damper is the one that has the same direction as
the damping torque τd at the point of application. It follows that the damping torque is given
by
 
τd ¼ b θ_ 2  θ_ 1 (3.3.10a)

for the end torques pointing away from each other as shown in Figure 3.3.6(b), and
 
τd ¼ b θ_ 1  θ_ 2 (3.3.10b)

for the end torques pointing toward each other as shown in Figure 3.3.6(c).
In some applications, one end of a viscous damper is fixed. Figure 3.3.7 shows two
examples of such dampers, one for three-dimensional bodies, and the other for slabs. In
both the examples, the torque–velocity relation is

τd ¼ bθ_ (3.3.11)
150 3 Mechanical Systems

.
θ2
.
θ1 lubricant

(a)

τd b τd τd b τd

. . . .
θ1 θ2 θ1 θ2

. . . .
(b) Δω = θ2 – θ1 (c) Δω = θ1 – θ2

Figure 3.3.6 Rotational viscous damper: (a) schematic; (b) Δω ¼ θ_ 2  θ_ 1 ; (c) Δω ¼ θ_ 1  θ_ 2

.
θ

.
θ
b τd
τd

(a) (b)

Figure 3.3.7 Rotational viscous damper with one end fixed: (a) damper for three-dimensional bodies; (b)
damper for slabs

In rotor machinery, bearings are used to support a rotating shaft and to reduce rotational
friction. A bearing can be viewed as a viscous damper in Figure 3.3.7(a). Due to its special
configuration and important applications, however, the working principle of bearings
deserves description. Without loss of generality, a schematic of a ball bearing is shown in
Figure 3.3.8(a), where the outer race of the bearing is stationary, the inner race is attached to
a rotating shaft, balls are used to separate the inner and outer races, and lubricant is added
between the races to minimize friction and prevent wear.
In the rotation of a bearing-supported shaft, viscous friction forces are induced by the
lubricant, ball–shaft contact, and other factors, as demonstrated in Figure 3.3.8(b), where fd
represents viscous friction forces acting on the shaft. These friction forces produce a resisting
damping torque τd that is described by Eq. (3.3.11). The symbols of bearings adopted in this
test are and , as shown in Figure 3.3.8(c). The damping torque by a bearing is always
against the shaft rotation as shown in the free-body diagram in Figure 3.3.8(d).
 3.3 Rotational Systems 151

Outer race

Inner race
Balls ω
Rotating
Lubricant shaft fd
ω

(a) (b)

Shaft Bearing
x

ω x Shaft
τd

Shaft Bearing ω

(c) (d)

Figure 3.3.8 Ball bearing: (a) schematic; (b) viscous friction forces acting on a rotating shaft; (c) notation
of bearing; (d) damping torque applied to the shaft by the bearing

The damping elements presented so far are linear. A nonlinear damping element is
considered as follows. For an object moving in the air, a drag force (air resistance) acts in
the opposite direction to the motion of the object, with an amplitude proportional to the
square of the velocity of the object relative to the air. Thus, the rotation of a body in the air (a
propeller for instance), due to an aerodynamic effect, is resisted by a drag torque, which can
be described by

τd ¼ bD θ_ sgnðθÞ
_
2
(3.3.12)

where the damping coefficient bD depends on drag coefficient, air mass density, and the shape
of the body. The notation of drag torque is shown in Figure 3.3.9.

Free-Body Diagrams and Auxiliary Plots

As a way of analysis (the third key in system modeling), free-body diagrams for mass
elements are drawn and an auxiliary plot for a spring or damping element with two movable
ends may be sketched. The development of a rotational system model follows the same steps
as listed in Section 3.2.3, provided that rotational elements are used.
152 3 Mechanical Systems

.
θ
τd

bD

Figure 3.3.9 Drag torque due to air resistance

Example 3.3.1
A rotor (motor–propeller assembly) of a quadcopter is shown in Figure 3.3.10, where the
propeller is driven by a torque τm of the electric motor, and it is subject to a drag torque due to
aerodynamic effect. The motor and the propeller are connected by a flexible shaft, which can be
modeled as a torsional spring of coefficient kT and the motor shaft is supported by a bearing of
coefficient b. The mass moment of inertia of the rotating parts of the motor is Jm and the mass
moment of inertia of the propeller is Jp . Let the rotation angles of the motor and propeller be θm
and θp , respectively. Derive the equations of motion for the rotor system.

bD
Propeller θp

Jp

Shaft Electric
motor
Bearing kT

θm

Jm b

τm

Figure 3.3.10 Simplified model of a motor–propeller assembly

Solution
The free-body diagrams of the motor and propeller and the auxiliary plot of the spring (flexible
shaft) are drawn in Figure 3.3.11, where τs , τb , and τD are the spring torque, bearing torque, and
drag torque, respectively. The application of Newton’s second law, Eq. (3.3.1), to the two mass
elements gives
 3.3 Rotational Systems 153

τm τb
Jm τs Jp τD
τs
θm θp

kT
τs τD

θm θp

Figure 3.3.11 Free-body diagrams and auxiliary plot of the motor–propeller assembly in Example 3.3.1

Motor : Jm θ€m ¼ τm  τb  τs (a)

Propeller : Jp θ€p ¼ τs  τD (b)

Also, the torques of the spring and damping elements are given by
 
τs ¼ kT Δθ ¼ kT θm  θp
(c)
τb ¼ bθ_m ; τD ¼ bD θ_ sgnðθ_p Þ
2
p

where, according to Figure 3.3.5(b), the relative rotation of the spring is Δθ ¼ θm  θp .

Substituting Eq. (c) into Eqs. (a) and (b) yields the equations of motion of the motor–propeller
assembly as follows
Jm θ€m þ bθ_ m þ kT θm  kT θp ¼ τm
(d)
Jp θ€p þ bD θ_ sgnðθ_p Þ þ kT θp  kT θm ¼ 0
2
p

3.3.2 Simple Rotational Systems

In this section, six simple rotational systems are modeled: the pulley, simple pendulum, lever,
slewing rigid bar, rotating shaft, and spinning rigid disk. These simple systems can be
combined with translational systems and rigid bodies to make complex mechanical systems,
as shall be shown in the subsequent sections.

Pulley
Pulleys are used to change the direction of movement of a taut cable (rope, belt, or
chain), or to transfer power between two elements that are connected by the cable. A
pulley is a cylinder that can rotate about its center, and that has a groove for a cable to
rest in it. A pulley is shown in Figure 3.3.12, where the center of the pulley is at fixed
hinge O and the cable is loaded by forces F1 and F2 . Assume that the cable is unextendible
154 3 Mechanical Systems

and that the rope does not slip on the surface of

the groove. The governing equation of motion for θ R
the pulley is given by
o

IO θ€ ¼ F1 R  F2 R (3.3.13)

where IO is the mass moment of inertia of the

pulley with respect to the hinge O. If the inertia F1 Io F2
of the pulley is negligible, IO ¼ 0, and the previous
equation yields F1 ¼ F2 . Such a pulley is called an Figure 3.3.12 Pulley under loads
ideal pulley.
In some applications, the center of a pulley is
movable. This renders the motion of the pulley in
both translation and rotation. This combination of translation and rotation is addressed in
Section 3.4.2.

Simple Pendulum
A simple pendulum is a lumped mass that is suspended by an unstretchable string at a fixed
point, as shown in Figure 3.3.13(a). With the gravity as a restoring force, the pendulum
sways freely about its equilibrium position ðθ ¼ 0Þ. The pendulum can be viewed as a
rotational system with the mass moment of inertia IO ¼ ml2 , where l is the string length.
The pendulum is subject to a moment MO ¼ mgl sin θ, where the negative sign means that
that the movement of gravity is always against the angular motion of the pendulum. By Eq.
(3.3.1),

ml2 θ€ ¼ mgl sin θ (3.3.14)

which leads to the nonlinear equation of motion of the pendulum

g
θ€ þ sin θ ¼ 0 (3.3.15)
l

The pendulum equation can also be obtained by treating the mass as a particle in a circular
motion. To this end, consider the free-body diagram in Figure 3.3.13(b). The application of
Newton’s second law, Eq. (3.1.2), in the tangential direction (perpendicular to the spring
tension T ) gives

maθ ¼ mg sin θ (3.3.16)

€ the same governing equation (3.3.15) for the

Because the tangential acceleration is aθ ¼ lθ,
simple pendulum is obtained.
 3.3 Rotational Systems 155

T
O

g m
l aθ

θ mg
m θ

(a) (b)

Figure 3.3.13 A simple pendulum: (a) schematic; (b) free-body diagram

If the angle θ of swing is small, sin θ ≈ θ and Eq. (3.3.15) is approximated by

g
θ€ þ θ ¼ 0 (3.3.17)
l

This linearized differential equation implies that the oscillation (back and forth swing
pffiffiffiffiffiffiffi
motion) of the pendulum has a natural period or simply period T ¼ 2π l=g, regardless of
the inertia or weight of the pendulum. Refer to Section 7.1.3 for the definition of the natural
period of a system in free oscillatory motion.

Lever
A lever is a rigid rod that is pivoted at a fixed hinge, and it is used to amplify or
transmit forces in machines and equipment. In Figure 3.3.14(a), a lever pivoted at point
O is constrained by a spring and is subject to a vertical force f. Assume that the
rotation angle θ of the lever is small such that the tip displacements of the lever can be
approximated by y1 ¼ aθ; y2 ¼ bθ. With the free-body diagram in Figure 3.3.14(b), use of
Eq. (3.3.1) gives

IO θ€ ¼ fa  fs b (3.3.18)

Because the spring force is fs ¼ ky2 ¼ kbθ, the equation of motion of the lever is

IO θ€ þ kb2 θ ¼ fa (3.3.19)

If the inertia of the lever is negligible, the applied force f and spring force are in equilibrium;
namely, kb2 θ ¼ fa.
156 3 Mechanical Systems

Io
y1 O

θ y2

a b

(a)
f

Ry

O
Rx

fs

(b)

Figure 3.3.14 A lever–spring system: (a) schematic; (b) free-body diagram

Slewing Rigid Bar

In Figure 3.3.15(a), a rigid bar rotates about a pivot point O under a torque τ. For a smooth
slewing motion, a bearing is installed at the pivot point. Figure 3.3.15(b) shows the free-body
diagram of the system, where τd is the damping torque applied by the bearing. It follows from
Eq. (3.3.1) that

IO θ€ ¼ τ  τd (3.3.20)

which with τd ¼ bθ_ leads to the governing equation of the slewing bar:

IO θ€ þ bθ_ ¼ τ (3.3.21)

The slewing bar is a model of single-link robotic arms.

 3.3 Rotational Systems 157

Io Io
τ τ
θ θ
O
Bearing, b τd O

(a) (b)

Figure 3.3.15 A slewing rigid bar: (a) schematic; (b) free-body diagram

kT θ
J x τ J
τb

x τs θ
τ
b
(a) (b)

Figure 3.3.16 A rotating shaft: (a) schematic; (b) free-body diagram

Rotating Shaft
Rotating shafts were considered in Example 3.3.1. This type of system is further discussed for
the preparation of modeling of geared systems in Section 3.3.4. In Figure 3.3.16(a), a torque τ
is applied to a bearing-supported rotating shaft that is constrained by a torsional spring. The
free-body diagram of the shaft is shown in Figure 3.3.16(b), where τs is the spring torque and
τb is the bearing torque. By Eq. (3.3.1),

J θ€ ¼ τ  τs  τb (3.3.22)

_ gives the equation of motion of the rotating shaft

which, by τs ¼ kT θ and τb ¼ bθ,

J θ€ þ bθ_ þ kT θ ¼ τ (3.3.23)

Spinning Rigid Disk

A spinning rigid disk can be viewed as a shortened version of a rotating shaft with or
without a bearing. This model is used for a rotating object with its dimension along
the rotating axis being relatively small and its mass moment of inertia being relatively
large. The governing equation of motion for a spinning disk can be obtained from
Eq. (3.3.1).
158 3 Mechanical Systems

3.3.3 Combination of Translational and Rotational Elements

In this section, mechanical systems composed of translational elements described in Section
3.2.1 and rotational elements described in Sections 3.3.1 and 3.3.2 are considered. Modeling
of this type of systems follows the same steps as described in Section 3.2.3, except that both
translational and rotational elements are involved.

Example 3.3.2
A pulley–mass system is shown in Figure 3.3.17(a), where an unextendible cable connects a
cylinder and a mass. Consider gravity. Derive a mathematical model of the system.
Solution
The free-body diagrams of the pulley and mass are shown in Figure 3.3.17(b), where T is the
tension force of the cable. Because the cable is unextendible, the rotation of the cylinder and the
displacement of the mass are related by y ¼ Rθ. Note that two free-body diagrams must be drawn
here even though the displacements of the cylinder and mass are not independent. By Newton’s
second law and its rotational version, the equations of the mass elements are

θ θ
Io Io
R R
O O
g

y
T

m
T
f f
y

mg
(a) (b)

Figure 3.3.17 A pulley–mass system with an unextendible cable: (a) schematic; (b) free-body
diagrams
 3.3 Rotational Systems 159

IO θ€ ¼ fR  TR
m€ y ¼ T  mg

It is easy to show that T ¼ mRθ€ þ mg and that by substitution

IO θ€ ¼ fR  ðmRθ€ þ mgÞR

This yields the governing equation of motion of the pulley–mass system

ðIO þ mR2 Þθ€ ¼ fR  mgR

Also, by y ¼ Rθ, an equivalent equation of motion in terms of y is obtained as follows


IO
þ m y€ ¼ f  mg
R2

In Example 3.3.2, the full description of the pulley–mass system motion only requires one
displacement parameter (either θ or y). If the cable connecting the cylinder and mass is
extendible, two independent displacement parameters are necessary; see Example 3.3.3.

Example 3.3.3
The pulley–mass system in Figure 3.3.18(a) has an extendible cable, which can be modeled as a
spring of coefficient k. Derive the equations of motion for the pulley-mass system.

θ θ fs
Io Io

R R Rθ
O O
g

fs y
k
y
fs
f m f
y fs

mg

(a) (b)

Figure 3.3.18 A pulley–mass system with an extendible cable: (a) schematic; (b) free-body diagrams
and auxiliary plot
160 3 Mechanical Systems

Solution
Because the cable is extendible, the rotation angle θ and displacement y are independent of each
other. In other words, y ≠ Rθ. Thus, these displacement parameters are all required to fully
describe the motion of the pulley–mass system. Figure 3.3.18(b) shows the free-body diagrams
of the pulley and mass and the auxiliary plot of the spring (cable). Thus, by Newton’s second law,

IO θ€ ¼ fR  fs R
m€ y ¼ fs  mg

With the spring force being fs ¼ kðRθ  yÞ, the equations of motion of the system are obtained as
follows

IO θ€ þ kR2 θ  kRy ¼ fR
m€y þ ky  kRθ ¼ mg

Degrees of Freedom
Comparison of Examples 3.3.2 and 3.3.3 shows that the minimum number of governing
equations of motion for a mechanical system is the number of independent displacement
parameters. This leads to the following definition about the degrees of freedom of mechan-
ical systems.

Definition: The number of degrees of freedom (DOFs) of a mechanical system is the

minimum number of independent displacement parameters needed to fully describe
the motion of the system. Also, the number of DOFs is equal to the number of
independent governing equations of motion.
According to the above definition, the pulley–mass system with an unextendible cable in
Example 3.3.2 is a one-DOF system; the pulley–mass system with an extendible cable in
Example 3.3.3 is a two-DOF system.
The concept of DOFs is applicable to all mechanical systems.

Example 3.3.4
In Figure 3.3.19(a), a cart of mass M carrying a simple pendulum with mass m and length l moves
on a smooth horizontal surface. The cart is constrained by a viscous damper (c) and is subject to
an external force f. For the pendulum to work, gravity (g) must be considered. For this two-DOF
system, derive the independent governing equations of motion. Also, obtain an expression of the
tension of the pendulum cable.
 3.3 Rotational Systems 161

x N T
x θ
c f
M
fd f
M
m B (xB, yB)
A
mg
Mg T y
g l
θ
θ x
m A
(a) (b)

Figure 3.3.19 A cart–pendulum system: (a) schematic; (b) free-body diagrams

Solution
The free-body diagrams of the cart and pendulum are shown in Figure 3.3.19(b), where T is
tension of the pendulum cable, N is the normal force by the horizontal surface, A is the point of the
cart at which the pendulum is suspended, Axy is the coordinate system with the origin at point A,
and xB and yB describe the location B of the pendulum mass. By Newton’s second law, the
displacement x of the cart is governed by

M x€ þ cx_ ¼ f þ T sin θ (a)

where fd ¼ cx_ has been used. Because the cart has no vertical motion, the forces applied to the cart
in the y direction are balanced: N ¼ Mg þ T cos θ. Also, the motion of the pendulum mass is
governed by

maBx ¼ T sin θ; maBy ¼ T cos θ  mg

where aBx and aBy are the acceleration components of mass m in the x and y directions, respect-
ively. The coordinates of the pendulum mass are

xB ¼ x þ l sin θ; yB ¼ l cos θ

by which
aBx ¼ x€B ¼ x€ þ lðθ€ cos θ  θ_ sin θÞ
2

aBy ¼ y€B ¼ lðθ€ sin θ þ θ_ cos θÞ

2

Thus, the governing equations of motion for the pendulum become

x þ lðθ€ cos θ  θ_ sin θÞ ¼ T sin θ
2
m½€ (b)

m½lðθ€ sin θ þ θ_ cos θÞ ¼ T cos θ  mg

2
(c)
162 3 Mechanical Systems

Of the three equations (a) to (c), only two are independent. This is because the cart–pendulum
system has only two independent displacement parameters (x and θ). The derivation of two
independent equations can be done by eliminating the tension force T. To this end, adding Eqs.
(a) and (b) gives

x þ cx_ þ mlðθ€ cos θ  θ_ sin θÞ ¼ f

2
ðM þ mÞ€ (d)

Also, multiplying Eq. (b) by cos θ, multiplying Eq. (c) by sin θ and adding the resulting equa-
tions, yields

x þ lðθ€ cos θ  θ_ sin θÞ cos θ þ m½lðθ€ sin θ þ θ_ cos θÞ sin θ þ mg sin θ ¼ 0
2 2
m½€

which is further reduced to

x€ þ lθ€ þ g sin θ ¼ 0 (e)

Equations (d) and (e) are two independent governing equations of motion for the cart–pendulum
system. Furthermore, multiplying Eq. (b) by sin θ, multiplying Eq. (c) by cos θ, and adding the
resulting equations, gives the expression of the cable tension as follows

T ¼ mðlθ_  x€ sin θ þ g cos θÞ

2
(f)

3.3.4 Geared Systems

Gears are rotating machine parts that are used to change the speed, torque, and
direction of rotation, and to transmit power in mechanical systems. Rotating systems
consisting of gears are called geared systems. There are several types of gears, including
spur, helical, bevel, crown, rack-and-pinion, and worm gears. In this section, spur gears
are considered. The approach in modeling and analysis can be applied to other types of
gears.

n1 Gear 1 n1
θ1 r1 θ1

θ1 θ2
r2
θ2
θ2
Gear 2
n2
n2

(a) (b) (c)

Figure 3.3.20 A pair of spur gears: (a) schematic; (b) side view, (c) notation
 3.3 Rotational Systems 163

Consider a pair of spur gears in Figure 3.3.20(a), where Gears 1 and 2 have n1 and n2 teeth,
respectively. Gear 1, which is mounted on a shaft with an applied toque τ, is called the input
gear or drive gear; Gear 2, which drives a connecting shaft, is called the output gear or driven
gear. In modeling, the following ideal gear assumptions are made:

(i) gears are massless; and

(ii) gears are in perfect meshing, without friction, backlash, and slipping.

According to these assumptions there is no energy loss in gear engagement.

Gears change speed, torque, and transmitting power through a gear ratio. A side view of
the spur gear pair is shown in Figure 3.3.20(b), where r1 and r2 are the radii of Gears 1 and 2
respectively. The gear ratio N of the gear pair (from the input gear to the output gear) is
defined as
n2 r2
N¼ ¼ (3.3.24)
n1 r1

The notation for a gear pair is shown in Figure 3.3.20(c).

There are several formulas about a gear pair in operation, which are presented as follows.

Kinematic Relation
Due to the ideal gear assumptions, the passage of the circumferential arc length of each gear
through a teeth engagement point must be the same. This means that

r1 θ1 ¼ r2 θ2 (3.3.25)

which can be written as

r 2 θ1
¼ ¼N (3.3.26)
r 1 θ2

Because of Eq. (3.3.25), rotational displacements θ1 and θ2 are not independent.

Torque Ratio
A pair of gears under torques are shown in Figure 3.3.21(a), where τ1 is a torque
applied to the drive gear by the connecting shaft; and τ2 is a resisting torque applied to the
driven gear by the connecting shaft. Torques τ1 and τ2 are always in the same direction
although they are applied to two different bodies. The free-body diagrams of the gears
are plotted in Figure 3.3.21(b), where F and N are the tangential and normal forces induced
by gear meshing. The equations of motion for the gears by Newton’s second law are as
follows
164 3 Mechanical Systems

Drive gear τ2 θ2
n1 τ1
τ1
r1
N N
θ1 θ2 r2

τ2 θ1
n2
Driven gear Gear 1 F F Gear 2

(a) (b)

Figure 3.3.21 Gears subject to torques: (a) schematic; (b) free-body diagrams

For Gear 1: I1 θ€1 ¼ τ1  Fr1

For Gear 2: I2 θ€2 ¼ Fr2  τ2

where I1 and I2 are the mass moments of inertia of the gears. Because the gears are massless,
I1 ¼ 0 and I2 ¼ 0. This renders zero resultant torque on each gear:

τ1 ¼ Fr1 ; τ2 ¼ Fr2 (3.3.27)

It follows that
τ2 r2
¼ ¼N (3.3.28)
τ1 r1

Thus, for a pair of ideal gears, its torque ratio is the same as the gear ratio N.

Work Done by Torques

By the ideal gear assumptions, the work done by torque τ1 through rotation θ1 should be the
same as the work done by torque τ2 through rotation θ2 :

τ1 θ1 ¼ τ2 θ2 (3.3.29)
τ2 θ1
This makes sense because ¼ ¼ N, as indicated by Eq. (3.3.28).
τ1 θ2
With the gear ratio N and the above-mentioned relevant formulas, governing equa-
tions of geared systems can be established. In a modeling process, an auxiliary plot for
a pair of gears as a whole can be generated, like Figure 3.3.21(a), and Eqs. (3.3.26) and
(3.3.28) can be used.

Example 3.3.5
Two bearing-supported rotating shafts are coupled by a pair of gears in Figure 3.3.22, where a
torque τ is applied to Shaft 1. This is a system with one DOF. Derive the equation of motion for the
geared system in θ2 . Also, obtain an equivalent equation of motion in θ1 .
 3.3 Rotational Systems 165

τ θ1
× n1
cT
J1 × θ2
× kT
×
× J2
b1
Shaft 1 ×
n2
Shaft 2 b2

Figure 3.3.22 Schematic of a geared system

Solution
The free-body diagrams of the shafts and the auxiliary plot of the gear pair are shown in Figure
3.3.23, where Tg1 and Tg2 are the torques applied to the gears by Shafts 1 and 2, respectively, like τ1
and τ2 in Figure 3.3.21(a). The equations of motion of the shafts are obtained as

For Shaft 1: J1 θ€1 ¼ τ  Tb1  Td  Tg1

For Shaft 2: J2 θ€2 ¼ Tg2  Tb2  Ts

The gear pair, by Eqs. (3.3.26) and (3.3.28), is described by

1
Tg2 ¼ NTg1 ; θ2 ¼ θ1
N

where N ¼ n2 =n1 is the gear ratio, and Eqs. (3.3.19) and (3.3.21) have been used. The internal
forces (torques) of the spring and damping elements are

For bearings: Tb1 ¼ b1 θ_ 1 ; Tb2 ¼ b2 θ_ 2

For torsional damper: Td ¼ cT θ_ 1

For torsional spring: Ts ¼ kT θ2

θ1
Tg
τ 1
J1
n1
Td Tb θ1
1
Shaft 1 Tg x
1

x Tg
θ2 2

n2
θ2 Tb
2
Gear pair
J2
Tg Ts
2
Shaft 2
(a) (b)

Figure 3.3.23 (a) Free-body diagrams and (b) auxiliary plot of the geared system in Example 3.3.5
166 3 Mechanical Systems

It follows that
J1 θ€1 ¼ τ  b1 θ_ 1  cT θ_ 1  Tg1
J2 θ€2 ¼ Tg2  b2 θ_ 2  kT θ2

The previous equations are not independent. To obtain an independent governing equation in
terms of θ2 , rewrite the previous equations by the gear equations, yielding

J1 N θ€2 ¼ τ  b1 N θ_ 2  cT N θ_ 2  Tg1 (a)

J2 θ€2 ¼ NTg1  b2 θ_ 2  kT θ2 (b)

From Eq. (a),

Tg1 ¼ J1 N θ€2  τ þ b1 N θ_ 2 þ cT N θ_ 2 (c)

Substituting Eq. (c) into Eq. (b) yields the equation of motion of the geared system in θ2:
   
J1 N 2 þ J2 θ€2 þ b1 N 2 þ cT N 2 þ b2 θ_ 2 þ kT θ2 ¼ Nτ (d)

Also, by θ2 ¼ θ1 =N, Eq. (d) can be reduced to an equivalent equation of motion in θ1

  
1 € 1 _ 1
J1 þ 2 J2 θ 1 þ b1 þ cT þ 2 b2 θ 1 þ kT θ1 ¼ τ
N N N2

For simplicity, the previous equation of motion can be written as

Je θ€1 þ be θ_ 1 þ ke θ1 ¼ τ
1 1 1
with effective parameters Je ¼ J1 þ 2
J2 , be ¼ b1 þ cT þ 2 b2 , and ke ¼ 2 kT .
N N N

Example 3.3.6
A motor drives a propeller through a pair of gears, as shown in Figure 3.3.24, where τm is the
motor torque; Jm and Jp are the mass moments of inertia of the motor–shaft assembly and the

τm
b1 n1

Jm θ2 b2
Jp

Motor θ1
bD
n2

Propeller

Figure 3.3.24 Schematic of a motor-gear-propeller system

 3.3 Rotational Systems 167

propeller–shaft assembly, respectively; b1 and b2 are the bearing coefficients, and θ1 and θ2 are the
rotational displacements of the motor shaft and propeller shaft, respectively. Assume that the
propeller is opposed by a drag torque due to an aerodynamic effect, which is characterized by
damping coefficient bD . Develop a mathematical model for the geared system.
Solution
This example is different from Example 3.3.1 in that the motor and propeller are coupled by a pair
of gears, instead of a flexible shaft. Thus, the system in Example 3.3.1 has two DOFs while the
geared system in the current example has only one DOF. The free-body diagrams of the mass
elements and the auxiliary plot of the gear pair are drawn in Figure 3.3.25, where Tg1 and Tg2 are
gear torques, and TD is a drag torque applied to the propeller. With the free-body diagrams in
Figure 3.3.25(a), the equations of motion of the system are written as

For motor  shaft assembly: Jm θ€1 ¼ τm  Tb1  Tg1 (a)

For propeller  shaft assembly: Jp θ€2 ¼ Tg2  Tb2  TD (b)

The internal forces of the damping elements are given by

For bearings: Tb1 ¼ b1 θ_1 ; Tb2 ¼ b2 θ_2 (c)

For aerodynamic effect: TD ¼ bD θ_p sgnðθ_p Þ

2
(d)

Plugging Eqs. (c) and (d) into Eqs. (a) and (b) yields the governing equations

Jm θ€1 þ b1 θ_ 1 ¼ τm  Tg1
(f)
Jp θ€2 þ b2 θ_ 2 þ bD θ_ sgnðθ_ p Þ ¼ Tg2
2
p

As discussed in Example 3.3.5, the previous two differential equations are not independent. By
the auxiliary plot in Figure 3.3.25(b), the gear equations are given by

1 n2
Tg2 ¼ NTg1 ; θ2 ¼ θ1 ; with N ¼ (g)
N n1
θ1
Tb n1
τm 1
Jm θ1
Tg
Tg 1
1

Motor–shaft assembly
θ2 Tg
θ2 2
Tb
Tg Jp 2
n2
2
TD
Gear pair
Propeller–shaft assembly
(a) (b)

Figure 3.3.25 The geared system in Example 3.3.6: (a) free-body diagrams; (b) auxiliary plot
168 3 Mechanical Systems

If θ2 is chosen as the independent displacement parameter, by substituting Eq. (g) into Eq. (f), the
independent governing equation of motion for the geared system is obtained as follows:
   
Jm N 2 þ Jp θ€2 þ b1 N 2 þ b2 θ_ 2 þ bD θ_ p sgnðθ_ 2 Þ ¼ Nτm
2
(h)

3.3.5 Transfer function Formulation

The establishment of transfer function formulations for rotational systems follows the same
steps for translational systems, as described in Section 3.2.4.

Example 3.3.7
Consider the motor–propeller assembly in Example 3.3.1. By ignoring drag torque, derive a
transfer function for the system, with the motor torque τm as the input and the rotation speed θ_ p
of the propeller as the output. Also, determine the order of the transfer function.
Solution
The governing equations (d) of motion in Example 3.3.1, without the drag torque, are

Jm θ€m þ bθ_ m þ kT θm  kT θp ¼ τm
(a)
Jp θ€p þ kT θp  kT θm ¼ 0

which are linear differential equations with constant coefficients. Taking Laplace transform of the
equations with respect to time and with zero initial values gives
 2 
Jm s þ bs þ kT Θm ðsÞ  kT Θp ðsÞ ¼ Τm ðsÞ
 2  (b)
Jp s þ kT Θp ðsÞ  kT Θm ðsÞ ¼ 0

where Θm ðsÞ, Θp ðsÞ, and Τm ðsÞ are the Laplace transforms of θm , θp and τm , respectively.
Elimination of Θm ðsÞ from Eq. (b) arrives at

  Jp s2 þ kT
Jm s2 þ bs þ kT Θp ðsÞ  kT Θp ðsÞ ¼ Τm ðsÞ (c)
kT

which leads to
 
s½ðJm s þ bÞ Jp s2 þ kT þ Jp kT sΘp ðsÞ ¼ kT Tm ðsÞ

By the definition given in Eq. (3.2.12), the transfer function of the motor–propeller system is
obtained as follows
sΘp ðsÞ skT
GðsÞ ¼ ¼   
Τm ðsÞ s ðJm s þ bÞ Jp s2 þ kT þ Jp kT s
kT
¼  
ðJm s þ bÞ Jp s2 þ kT þ Jp kT s

This is a third-order transfer function.

 3.4 Rigid-Body Systems in Plane Motion 169

3.4 Rigid-Body Systems in Plane Motion

The mechanical systems modeled so far are composed of those mass elements which are in
either translation or rotation about a fixed axis. Considered in this section are the two-
dimensional motion of slab-like rigid bodies, in a combination of translation and rotation.
Description of three-dimensional motion of rigid bodies, which is much more involved, is
beyond the scope of this text.

3.4.1 Three Keys in System Modeling

As before, modeling of this type of mechanical system emphasizes three keys: fundamental
principles, basic elements, and free-body and auxiliary plots.

Mass Center and Relative Motion

Consider a rigid body contained in domain Ω; see Figure 3.4.1. The mass center G of the body
(also called the center of gravity) is defined by
ð
1
rG ¼ r dm (3.4.1)
m
Ω

ð
where m ¼ dm is the mass of the body, and r is the position vector of the differential mass
Ω
dm. In Cartesian coordinates, the position of the mass center is given by
ð ð
1 1
xG ¼ x dm; yG ¼ y dm (3.4.2)
m m
Ω Ω

G
dm
rG
r
y

O x

Figure 3.4.1 The mass center of a rigid body

170 3 Mechanical Systems

A general plane motion of a rigid body can be decomposed into a translation defined by
the motion of its mass center G and a simultaneous rotation about G. As shown in Figure
3.4.2, the velocity vP of any point P on the rigid body is the sum of the velocity vG of G and the
velocity vP=G relative to G:

vP ¼ vG þ vP=G (3.4.3)

The velocity vP=G is a relative velocity vector associated with rotation about the reference
point G and it can be expressed by

vP=G ¼ r_ P=G ¼ ωk  rP=G (3.4.4)

where ω is the angular velocity of the body, which is independent of the reference point, and k
is a unit vector perpendicular to plane Oxy, which is determined by the right-hand rule. The
acceleration of point P is

aP ¼ aG þ aP=G (3.4.5)

where aG is the velocity of G; aP=G is a relative acceleration associated with rotation about the
reference point G and it is given by

aP=G ¼ v_ P=G ¼ αk  rP=G  ω2 rP=G (3.4.6)

with α being the angular acceleration of the body ðα ¼ ω_ Þ.

The above description of relative motion is applicable to any reference point selected on a
rigid body. For instance, for a reference point A, the velocity and acceleration of point B are
determined by Eqs. (3.4.3)–(3.4.6) with G and P replaced by A and B.

vp
vG
vP/G P
P P
rP/G
vG = vG +
ω
G G G
Ω Ω Ω

Plane motion Translation with G Rotation about G

Figure 3.4.2 Plane motion as the sum of translation and rotation

 3.4 Rigid-Body Systems in Plane Motion 171

Example 3.4.1
In Figure 3.4.3, a disk is rolling on a horizontal surface with rotation speed ω. There is no slipping
at point B so that the velocity at the point of contact is zero. If the center C of the disk is taken as a
reference point, the motion of the disk is the sum of the translation with C and rotation about C.
For instance, the velocity at B is vB ¼ vC  Rω. Because vB ¼ 0, vC ¼ Rω. Point B with zero
velocity is referred to as an instantaneous center of the disk. Also, by choosing B as a reference
point, the velocity magnitudes at A, C, and D are determined as follows

vA ¼ jABj ω ¼ 2R ω; vC ¼ jCBjω ¼ R ω; vD ¼ jDBj ω

A
vA

D
vD

C ω
vC

vB = 0

Figure 3.4.3 A disk rolling on a horizontal surface without slipping

As in the modeling of translational and rotational systems, the development of mathem-

atical models for rigid-body systems has the following three keys.

Key 1: Fundamental Principle

The plane motion of a slab-like rigid body, by Newton’s second law, is governed by the
following three differential equations about the mass center G of the body
X
Fx ¼ maGx ¼ m€xG
X
Fy ¼ maGy ¼ m€yG (3.4.7)
X
MG ¼ IG α ¼ IG θ€

where xG and yG are the displacement components of G, θ is the angular displacement of the
X X
body, m is the mass of the body, IG is the mass moment of inertia about G, Fx and Fy are
172 3 Mechanical Systems

X
the resultants of forces acting on the body, and MG is the resultant of force moments and
torques about G.

Key 2: Basic Elements

Mass Elements A slab-like rigid body of arbitrary shape is a mass element if its inertia
ðm; IG Þ is nonnegligible. For such an element, the equations of motion in Eq. (3.4.7) can be
applied.

Lumped Spring and Damping Elements The spring elements and damping elements (trans-
lational and torsional springs, translational and torsional dampers, bearings and aero-
dynamic drag, and flexible bodies with effective spring coefficients) can all be used in the
modeling of rigid-body systems. For these elements, the same force–displacement relations,
as described in Sections 3.2 and 3.3, are valid.

Key 3: Free-Body Diagrams and Auxiliary Plots

In the modeling of a rigid-body system, a free-body diagram for each mass element is drawn.
Also, auxiliary plots for spring and damping elements can be generated if needed.
With the above-mentioned three keys, the system modeling takes the same four steps as
described in Section 3.2.3 provided that Eq. (3.4.7) is applied to rigid-body mass elements.

3.4.2 Examples of Rigid-Body Systems

In this section, pulleys or disks with movable rotation centers, wheels in rolling and sliding
motion, and a combination of rigid bodies and lumped masses are examined. The equations
of motion for these systems shall be derived based on the process described in Section 3.4.1.

Pulleys or Disks with Movable Pins

A pulley with fixed pin (rotation center) was considered in Section 3.3.2. If the pin of a pulley
or disk is movable, the motion of the rigid body involves both translation and rotation.

Example 3.4.2
Consider a uniform disk in Figure 3.4.4(a), which is suspended by a spring at its rotation center
(also the mass center G). Under forces F1 , F2 and gravity, the disk experiences translation x and
rotation θ, which are independent of each other. The free-body diagram of the disk is shown in
Figure 3.4.4(b). The application of Eq. (3.4.7) gives
 3.4 Rigid-Body Systems in Plane Motion 173

x þ kx ¼ mg þ F1 þ F2
m€
IG θ€ ¼ F1 R  F2 R

where fs ¼ kx has been used. In this example, the disk has no horizontal motion.
If the inertia of the disk is negligible ðm ¼ 0; IG ¼ 0Þ, the previous dynamic equations become
the following equilibrium equations

kx ¼ mg þ F1 þ F2
F1 ¼ F 2

fs
k

m, IG
θ θ
g
R R

G G

x x

mg
F1 F2 F1 F2
(a) (b)

Figure 3.4.4 A disk with movable rotation center: (a) schematic and (b) free-body diagram

Wheel in Rolling and Sliding Motion

A wheel on a flat surface can involve both rolling and sliding motions.

Example 3.4.3
In Figure 3.4.5(a), a uniform wheel moves on a horizontal surface and it is subject to gravity and a
horizontal force F. Here, x is the displacement of the mass center G, and θ is the rotational
displacement of the wheel. The free-body diagram of the wheel is shown in Figure 3.4.5(b), where
fμ is the friction force between the wheel and surface at the contact point C. The motion of the
wheel has the following three possibilities.
(i) Pure Sliding Motion
If there is no friction ð fμ ¼ 0Þ, the wheel is in a translational motion with displacement x, just like a
lumped mass. The motion of the wheel is governed by

x¼F
m€
174 3 Mechanical Systems

x x

θ θ

G F G F
g R
mg
C
fμ
C

N
(a) (b)

Figure 3.4.5 A disk with movable rotation center: (a) schematic and (b) free-body diagram

(ii) Nonslipping Motion

Friction between the wheel and the surface exits, but the wheel does not slip on the surface. The
nonslipping condition is that the velocity of the wheel at the contact point C is zero:

vC ¼ x_  Rθ ¼ 0

In this case, the wheel is in both translation and rotation, as described by the following equations

m€x ¼ F  fμ
IG θ€ ¼ fμ R

where the friction force fμ ≤ μS N, with μS being a coefficient of static friction. Because x ¼ Rθ, only
one of the displacement parameters (x and θ) is independent. Through elimination of the friction
force in the previous equations, an independent equation of motion of the wheel is

ðIG þ mR2 Þθ€ ¼ FR

or

1
mþ IG x€ ¼ F
R2

(iii) Mixed Rolling and Slipping Motion

For wheel slips on a surface that has friction, the governing equations of motion are given by

_
x ¼ F  μD NsgnðxÞ
m€

IG θ€ ¼ μD NsgnðxÞR
_
 3.4 Rigid-Body Systems in Plane Motion 175

where Eq. (3.2.11) for friction force has been used. The coefficient μD of kinetic friction is usually
smaller than the coefficient μS of static friction. In this case, with x ≠ Rθ, x and θ are the
independent displacement parameters, which are needed to completely describe the mixed rolling
and sliding motion of the wheel.

Combination of Rigid Bodies and Lumped Masses

Models of mechanical systems often consist of both rigid bodies and lumped-mass elements.
See Example 3.4.4.

Example 3.4.4
In Figure 3.4.6, a cart carrying a uniform rigid bar moves on a smooth horizontal surface. One end
of the bar with mass m and length l is hinged at the point O of the cart with mass M, which is
constrained by a viscous damper of coefficient c. Under gravity, the rigid bar is an inverted
pendulum. This system has two independent displacement parameters: x and θ. Derive the
governing equations of motion of the cart–pendulum system.
Solution
The free-body diagrams of the bar and cart are drawn in Figure 3.4.7, where Rx and Ry are the
components of the reaction force at the hinge O, and N is the normal force provided by the surface.
By Eq. (3.4.7), the motion of the cart and pendulum is governed by

Cart M x€ ¼ f  fd  Rx
Pendulum maGx ¼ Rx ; maGy ¼ Ry  mg
l l
IG θ€ ¼ Ry sin θ  Rx cos θ
2 2
where the damping force fd ¼ cx, _ and aGx and aGy are the components of the acceleration of the bar
at its mass center G, in the x- and y-directions. The coordinates of the mass center can be written as

θ m, l
g

G
M x
c

f
O

Figure 3.4.6 A cart carrying an inverted pendulum

176 3 Mechanical Systems

N m, l
θ

x G
Rx
M f mg
fd y Rx
Ry Mg O

x Ry

Figure 3.4.7 Free-body diagrams of the cart–pendulum system in Example 3.4.4

l l
xG ¼ x þ sin θ; yG ¼ cos θ
2 2
This yields
l 
aGx ¼ x€G ¼ x€ þ θ€ cos θ  θ_ sin θ
2
2
l € 
yG ¼  θ sin θ þ θ_ cos θ
2
2

It follows that the equations of motion for the system are

M x€ þ cx_ ¼ f  Rx (a)
l 
m x€ þ θ€ cos θ  θ_ sin θ ¼ Rx
2
(b)
2
l
m ½θ€ sin θ þ θ_ cos θ ¼ Ry  mg
2
(c)
2
l l
IG θ€ ¼ Ry sin θ  Rx cos θ (d)
2 2
Because the system has two independent displacement parameters (x, θ), the previous four
equations are not independent. By Eq. (c)
l
Ry ¼ mg  m ½θ€ sin θ þ θ_ cos θ
2
(e)
2

Substituting Eqs. (b) and (e) into Eqs. (a) and (d) gives two independent equations of motion as
follows:

l € 
θ cos θ  θ_ sin θ ¼ f
2
x þ cx_ þ m
ðM þ mÞ€
2 (f)
ml2 € l l
θ þ m€
x cos θ  mg sin θ ¼ 0
3 2 2
 3.5 Energy Approach 177

1 2 1
where IG ¼ ml for the uniform bar has been used. The ml2 in the second equation is the mass
12 3
moment of inertia of the bar about the hinge point O, which can be obtained by the parallel-axis
theorem given in Section 3.3.1.

3.5 Energy Approach

In the previous sections, the equations of motion for a mechanical system are derived
through use of Newton’s laws to the free-body diagrams. This way of system modeling is
called the Newtonian approach. In this section, an alternative method, which makes use of
energy functions and produces a set of differential equations of motion called Lagrange’s
equations, is introduced. This method is referred to as the Lagrangian approach or energy
approach. As shown in Section 3.1.4, spatial integration of Newton’s second law leads to
the principle of work and energy. The Lagrangian approach follows this principle, and
hence is completely compatible with the Newtonian approach. The Lagrangian approach
in general is more capable of modeling complex mechanical systems with multiple
components.

3.5.1 Energy Functions of Mechanical Systems

In this section, kinetic energy, potential energy, and the Rayleigh dissipation function are
presented to prepare for the derivation of Lagrange’s equations in the next section.

Kinetic Energy
As shown in Section 3.1, the kinetic energy of a particle is

1 1
T ¼ mv  v ¼ mv2 (3.5.1)
2 2

For a rotating body about a fixed axis or pin O, as described in Section 3.3, its kinetic
energy is

1
T ¼ IO θ_
2
(3.5.1)
2

For a rigid body in plane motion, as shown in Figure 3.5.1, its kinetic energy is given by

1 1
T ¼ mv2G þ IG θ_
2
(3.5.2)
2 2

where vG ¼ jvG j, with vG being the velocity of the body at its mass center G.
178 3 Mechanical Systems

vG

G
θ

m, IG

Figure 3.5.1 A rigid body in plane motion

Potential Energy
The potential energy due to gravity and elastic deformation of a spring with a fixed end, as
discussed in Section 3.1.4, has the form

Gravity: V ¼ mgy (3.5.3)

1
Spring force: V ¼ kx2 (3.5.4)
2

where y is a coordinate that is in the opposite direction to the gravitational acceleration.

For a spring with two movable ends, as shown in Figure 3.2.2(b, c), its potential energy is
given by

1
V ¼ kðx1  x2 Þ2 (3.5.5)
2

Rayleigh Dissipation Function for Viscous Dampers

For a viscous damper with a fixed end, the Rayleigh dissipation function is

1
R ¼ cx_ 2 (3.5.6)
2

For a viscous damper with two movable ends, as shown in Figure 3.2.4(b, c), the Rayleigh
dissipation function is given by

1
R ¼ cðx_ 1  x_ 2 Þ2 (3.5.7)
2

Note that a Rayleigh dissipation function is in quadratic form.

 3.5 Energy Approach 179

Example 3.5.1
For the quarter-car suspension model in Figure 3.2.13, the energy functions are as follows:

1 1
T ¼ m1 y_ 21 þ m2 y_ 22
2 2
1 1
V ¼ k1 ðy1  y0 Þ þ k2 ðy2  y1 Þ2 þ m1 gy1 þ m2 gy2
2
2 2
1
R ¼ cðy_ 2  y_ 1 Þ2
2

3.5.2 Lagrange’s Equations

The establishment of Lagrange’s equations requires the usage of a minimum number of
displacement parameters or coordinates that are independent and that can fully describe the
motion of a mechanical system in consideration. This number is called the number of DOFs
of the system, as has been defined in Section 3.3.3.
Let q1 ; q2 ; . . . ; qn be a set of n independent coordinates for a mechanical system. The
system thus has n degrees of freedom (DOFs) and it is thus called an n-DOF system. The
energy functions for the system can be written as

T ¼ Tðq1 ; q2 ; . . . ; qn ; q_ 1 ; q_ 2 ; . . . ; q_ n Þ
V ¼ Vðq1 ; q2 ; . . . ; qn Þ (3.5.8)
R ¼ Tðq_ 1 ; q_ 2 ; . . . ; q_ n Þ

Note that kinetic energy T can be a function of q1 ; q2 ; . . . ; qn due to nonlinearities. With

these energy functions, Lagrange’s equations read

d ∂T ∂T ∂R ∂V
 þ þ ¼ Qi ; i ¼ 1; 2; . . . ; n (3.5.9)
dt ∂q_ i ∂qi ∂q_ i ∂qi

where Qi is the resultant force associated with coordinate qi , and it is called a generalized
force. The force Qi consists of external forces and friction forces, but it must exclude
gravitational forces and forces from springs and viscous dampers because they have been
considered in the energy functions V and R. For this reason, Qi is a nonconservative force. In
the application of Lagrange’s equations, free-body diagrams are not needed.
Generalized forces can be identified from the following incremental work

X
n
dWnc ¼ Qj dqj (3.5.10)
j¼1

Rigorously speaking, the incremental work should be written as

180 3 Mechanical Systems

X
n
δWnc ¼ Qj δqj
j¼1

where δ is a variation operator, δWnc is called virtual work, and δqj is the virtual displace-
ment. Due to the limited mathematical background covered in this undergraduate text,
Eq. (3.5.10) is used instead.

Example 3.5.2
Consider again the quarter-car suspension model in Figure 3.2.13, which is described by two
independent coordinates, y1 and y2 . The energy functions of the system were obtained in Example
3.5.1. By Eqs. (3.5.9), the equations of motion for the system are derived as follows:
for y1
d
ðm1 y_ 1 Þ  cðy_ 2  y_ 1 Þ þ k1 ðy1  y0 Þ  k2 ðy2  y1 Þ þ m1 g ¼ 0
dt

which yields
m1 y€1 þ cy_ 1 þ ðk1 þ k2 Þy1  cy_ 2  k2 y2 ¼ m1 g þ k1 y0 (a)

and for y2
d
ðm2 y_ 2 Þ þ cðy_ 2  y_ 1 Þ þ þk2 ðy2  y1 Þ þ m2 g ¼ 0
dt

which yields
m2 y€2 þ cy_ 2 þ k2 y2  cy_ 1  k2 y1 ¼ m2 g (b)

Equations (a) and (b) are the same as those obtained by the Newtonian approach in Example 3.2.4.

Example 3.5.3
For the cart–pendulum system in Example 3.4.4, derive the governing equations of motion by the
Lagrange approach.
Solution
With Figure 3.4.6, the energy functions of the system are obtained as

1 1 1 1 2 _2
T ¼ M x_ þ mðvGx þ vGy Þ þ
2 2 2
ml θ
2 2 2 12
l 1
V ¼ mg cos θ; R ¼ cx_ 2
2 2
where vGx and vGy are the components of the velocity of the bar at its mass center G in the x- and
1 l l
y-directions, and IG ¼ ml2 has been used. Because xG ¼ x þ sin θ and yG ¼ cos θ,
12 2 2
 3.6 Block Diagrams of Mechanical Systems 181

l
vGx ¼ x_ G ¼ x_ þ θ_ cos θ
2
l
vGy ¼ y_ G ¼  θ_ sin θ
2

Hence, the kinetic energy is of the form

"  # 
2 2
1 1 l l 1 1 2 _2
T ¼ M x_ þ m x_ þ θ_ cos θ þ  θ_ sin θ
2
þ ml θ
2 2 2 2 2 12

1 1 1 2 _2 1
¼ ðM þ mÞx_ þ 2
ml θ þ ml x_ θ_ cos θ
2 2 3 2

According to Eq. (3.5.10), the incremental work by generalized forces is dWnc ¼ fdx, indicating
that Qx ¼ f and Qθ ¼ 0.
By Eq. (3.5.9), the equations of motion of the cart-pendulum system are obtained as follows:
for x,

d 1
ðM þ mÞx_ þ ml θ_ cos θ þ cx_ ¼ f
dt 2

yielding
l € 
θ cos θ  θ_ sin θ ¼ f
2
x þ cx_ þ m
ðM þ mÞ€ (a)
2
and for θ,

d 1 2_ 1 1 l
ml θ þ ml x_ cos θ þ ml x_ θ_ sin θ þ mg sin θ ¼ 0
dt 3 2 2 2

yielding
ml2 € l l
θ þ m€
x cos θ  mg sin θ ¼ 0 (b)
3 2 2

Note that Eqs. (a) and (b) are the same as Eq. (f) in Example 3.4.4. The latter is obtained after the
elimination of the hinge reaction forces Nx and Ny . This example shows that the Lagrangian
approach automatically delivers independent equations of motion.

3.6 Block Diagrams of Mechanical Systems

The transfer function formulation described in Sections 3.2 and 3.3 is one form of model
representation for modeling and analysis of linear time-invariant mechanical systems. In this
section, we introduce another model representation, namely, the block diagram. Block-
diagram representation is a useful tool for modeling and analysis of mechanical systems.
Block diagrams have been widely used in the system response analysis of dynamic systems
182 3 Mechanical Systems

(Chapter 7) and the design of feedback control systems (Chapter 8). There are two types of
block diagrams: s-domain block diagrams and time-domain block diagrams. This section
concerns s-domain block diagrams.
The concept of s-domain block diagrams was introduced in Section 2.7. There are four
basic components of block diagrams: block, signal, summing point, and pick-off point. In
addition, for complicated diagrams, a crossing bridge is often used. These components are
listed in Table 3.6.1. A block diagram for a dynamic system is constructed through assembly
of these basic components. Like in a transfer function formulation, Laplace transform of the
governing equations of a system with zero initial conditions is performed in construction of a
block diagram. Block diagrams in the s-domain, as suggested by its name, are only valid for
linear time-invariant dynamic systems.
Construction of a block diagram for a mechanical system takes the following four steps.

(1) Derive the equations of motion for the system.

(2) Take the Laplace transform of the equations of motion with zero initial conditions.

Table 3.6.1 Basic components of the s-domain block diagram

Component Symbol Rule

Block Y ðsÞ ¼ GðsÞRðsÞ

Signal Variable, input or output

Summing point W ðsÞ ¼ U ðsÞ  V ðsÞ

Z ðsÞ ¼ X ðsÞ þ V ðsÞ

Pick-off point V ðsÞ ¼ U ðsÞ

X ðsÞ ¼ U ðsÞ

V (s)
Crossing bridge (without connection) Signals U(s) and V(s) do not interfere
with each other
U (s)
 3.6 Block Diagrams of Mechanical Systems 183

(3) Create parts of the block diagram with the Laplace transform equations and by the basic
components in Table 3.6.1.
(4) Assembly the parts obtained in Step 3 and use the preassigned inputs and outputs, to
complete the block diagram for the system.

These steps are demonstrated in Example 3.6.1.

Example 3.6.1
In this example, we explain the construction of the block diagram for the mechanical system
shown in Figure 1.3.2 in Chapter 1. For this system shown, its equations of motion are given in
Eq. (1.3.12), namely,

m1 x€1 þ cx_ 1 þ ðk þ k1 Þx1  k1 x2 ¼ 0 (a)

m2 x€2 þ k1 x2  k1 x1 ¼ f (b)

Consider a block diagram of the system with external force f as the input and
displacement x1 as the output. By Laplace transform and rearrangement, Eqs. (i) and (ii) are
reduced to

1
X1 ðsÞ ¼ k1 ðX2 ðsÞ  X1 ðsÞÞ (c)
m1 þ cs þ ks2
1 n o
X2 ðsÞ ¼ FðsÞ  k 1 ðX 2 ðsÞ  X 1 ðsÞÞ (d)
m2 s2

With the basic components in Table 3.6.1, Eqs. (c) and (d) are represented by the parts in Figure
3.6.1. Assembly of these parts by using summing point and pick-off point eventually yields the
block diagram in Figure 3.6.2, where Fs ðsÞ ¼ k1 ðX2 ðsÞ  X1 ðsÞÞ. The block diagram is the same as
in Figure 1.3.3.

k1 (X2 (s) – X1 (s)) 1 X1 (s) F (s) 1 X2 (s)

+
m1 s2 + cs + k m2s 2
–
k1 (X2 (s) – X1 (s))

Figure 3.6.1 Parts of the block diagram in Example 3.6.1: (a) representation of Eq. (c);
(b) representation of Eq. (d)
184 3 Mechanical Systems

F (s) – 1 X2 (s) 1 X1 (s)

+ + k1
m2 s2 – m1 s2 + cs + k
Fs (s)

Figure 3.6.2 Block diagram of the mechanical system in Figure 1.3.2

Example 3.6.2
Consider the spring–mass–damper system in Figure 3.2.8(a). Let the system inputs be f1 and
f2 . Let the system outputs be x_ 1 and x2 . Draw a block diagram for the system.
Solution
From Example 3.2.1, the governing equations of motion of the system are

m1 x€1 ¼ f1 þ k2 ðx2  x1 Þ  k1 x  cx_ 1 (a)

m2 x€2 ¼ f2  k2 ðx2  x1 Þ (b)

Performing Laplace transform of Eqs. (a) and (b) with zero initial conditions gives
 
m1 s2 þ cs þ k1 X1 ðsÞ ¼ F1 ðsÞ þ k2 ðX2 ðsÞ  X1 ðsÞÞ
m2 s2 X2 ðsÞ ¼ F2 ðsÞ  k2 ðX2 ðsÞ  X1 ðsÞÞ

or
1 n o
X1 ðsÞ ¼ F 1 ðsÞ þ k ðX
2 2 ðsÞ  X 1 ðsÞÞ (c)
m1 s2 þ cs þ k1
1 n o
X2 ðsÞ ¼ F 2 ðsÞ  k 2 ðX 2 ðsÞ  X 1 ðsÞÞ (d)
m2 s2

With the basic components in Table 3.6.1, (c) and (d) are represented by the parts in
Figure 3.6.3.
Assembling the parts in Figure 3.6.3 gives the block diagram of the spring–mass–damper system
in Figure 3.6.4, where V1 ðsÞ ¼ L½x_ 1  ¼ sX1 ðsÞ and Fs ðsÞ represents the internal force of the spring
ðk2 Þ connecting the two masses.
 3.7 State-Space Representations 185

Figure 3.6.3 Parts of the block diagram in Example 3.6.2: (a) representation of Eq. (c); and (b) repre-
sentation of Eq. (d)

Figure 3.6.4 A block diagram of the mechanical system in Example 3.6.2

3.7 State-Space Representations

Transfer function formulations and s-domain block diagrams are only valid for linear time-
invariant systems. On the other hand, linear time-variant systems and nonlinear systems,
such as the rocket in Example 3.2.6 and the coupled motor–propeller system in Example
3.3.1, are seen in many engineering applications. For these dynamic systems, modeling,
186 3 Mechanical Systems

Table 3.7.1. Comparison of state-space representations and transfer function formulations

State-space representation Transfer function formulation

System Linear or nonlinear Linear and time-invariant

Time-invariant or time-variant
Domain Time domain Laplace transform domain
Format State equations and output equations Input–output relation via transfer function

simulation, and analysis rely on other forms of model representation. In this section,
state-space representations for general dynamic systems, including time-variant and
nonlinear mechanical systems, are introduced. Also, as shall be seen in Section 6.5,
block diagrams in the time domain are another tool for modeling and simulation of
general dynamic systems.
A state-space representation, which is also called a state-space model, consists of two sets
of equations in the time domain: state equations and output equations. State equations are a
set of coupled first-order differential equations about quantities known as state variables.
Output equations are a set of coupled algebraic equations about the state variables. State
variables for a dynamic system are independent quantities that can completely describe the
future behavior of the system with known inputs and initial disturbances. State equations
and output equations can be either linear or nonlinear. A comparison for state-space
representations and transfer function formulations is given in Table 3.7.1.
Suppose that a system has n state variables x1 ; x2 ; . . . ; xn , m inputs u1 ; u2 ; . . . ; um and p
outputs y1 ; y2 ; . . . ; yn , all of which in general are functions of time. If the system is nonlinear,
its state equations can be written as
8
>
> x_ 1 ¼ f1 ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; um Þ
>
<x_ 2 ¼ f2 ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; um Þ
.. (3.7.1)
>
> .
>
:
x_ n ¼ fn ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; um Þ

and its output equations are of the form

8
>
> y1 ¼ g1 ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; um Þ
>
<y2 ¼ g2 ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; um Þ
.. (3.7.2)
>
> .
>
:
yp ¼ gp ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; um Þ

where fi and gj are nonlinear functions of the state variables and inputs. Note that in a state
equation, the left-hand side only contains the first derivative of a state variable and the right-
hand side does not have any derivative term of xk or ul .
 3.7 State-Space Representations 187

If the system is linear, which can be either time-invariant or time-variant, its state
equations and output equations are given by
8
> x_ 1 ¼ a11 x1 þ a12 x2 þ . . . þ a1n xn þ b11 u1 þ b12 u2 þ . . . þ b1m um
>
<x_ 2 ¼ a21 x1 þ a22 x2 þ . . . þ a2n xn þ b21 u1 þ b22 u2 þ . . . þ b2m um
.. (3.7.3)
>
> .
:
x_ n ¼ an1 x1 þ an2 x2 þ . . . þ ann xn þ bn1 u1 þ bn2 u2 þ . . . þ bnm um

and
8
>
> y1 ¼ c11 x1 þ c12 x2 þ . . . þ c1n xn þ d11 u1 þ d12 u2 þ . . . þ d1m um
>y ¼ c x þ a x þ . . . þ c x þ d u þ d u þ . . . þ d u
< 2 21 1 22 2 2n n 21 1 22 2 2m m
.. (3.7.4)
>
> .
>
:
yp ¼ cp1 x1 þ cp2 x2 þ . . . þ cpn xn þ dp1 u1 þ dp2 u2 þ . . . þ dpm um

where aij , bjk , ckl , and dlm are either constants or known functions of t. The linear state
equations and output equations can be cast in a matrix form as follows:

State equation: x_ ðtÞ ¼ AxðtÞ þ BuðtÞ (3.7.5)

Output equation: yðtÞ ¼ CxðtÞ þ DuðtÞ (3.7.6)

where xðtÞ, uðtÞ, and yðtÞ are the state, input, and output vectors, respectively, given by
0 1 0 1 0 1
x1 ðtÞ u1 ðtÞ y1 ðtÞ
Bx2 ðtÞ C Bu2 ðtÞ C By2 ðtÞ C
B C B C B C
xðtÞ ¼ B . C; uðtÞ ¼ B . C; yðtÞ ¼ B . C (3.7.7)
@ .. A @ .. A @ .. A
xn ðtÞ up ðtÞ yq ðtÞ

and A, B, C, and D are matrices composed of aij , bjk , ckl , and dlm , respectively.
Assume that a mathematical model for the mechanical system in consideration has been
obtained by following the guidelines in Section 3.2.3. Conversion of the model to a state
space representation takes the following four steps.

(1) Specify inputs and outputs of the system. The inputs are naturally the disturbances to the
system, including external forces and prescribed displacements. The outputs, depending
on the application and interest, can be displacements, velocities, accelerations, and
internal forces of the spring and damping elements.
(2) Select state variables. For a mechanical system, its displacements and velocities can be
conveniently chosen as state variables. The selection of state variables, however, is not
unique. Regardless of how state variables are selected, the total number of state variables
should be the same and it is the sum of the orders of the governing differential equations of
the system model. Refer to Section 2.5.4 for some examples of the selection of state
variables.
188 3 Mechanical Systems

(3) Differentiate the state variables one by one and perform mathematical manipula-
tions to obtain state equations of the form of Eq. (3.7.1) or Eq. (3.7.3). In this
process, the definition of the state variables and the governing differential equations
are used.
(4) Derive output equations of the form of Eq. (3.7.2) or Eq. (3.7.4), by the specified outputs
and selected state variables.

Example 3.7.1
For the mass on a slope in Figure 3.2.11, establish a state-space representation, with the gravita-
tional force as the input and the velocity of the mass and spring force as the outputs.
Solution
From Example 3.2.3, the equation of motion for the mass is
::
_ ¼ F þ Mg sin α
m x þkx þ μD Mg cos α sgnðxÞ (a)

which is a second-order nonlinear differential equation. Accordingly, two state variables are
selected as below
x1 ¼ x; x2 ¼ x_ (b)

By differentiating the first state variable, x_ 1 ¼ x,

_ the first state equation is obtained as

x_ 1 ¼ x2 (c)

Differentiate the second state variable, x_ 2 ¼ x€, and substitute the result into Eq. (a), to obtain

mx_ 2 þ kx1 þ μD Mg cos α sgnðx2 Þ ¼ F þ Mg sin α (d)

where Eq. (b) has been used. Thus, the second state equation is

1
x_ 2 ¼ ½kx1  μD Mg cos α sgnðx2 Þ þ F þ Mg sin α (e)
m

The outputs of the system are

_ y2 ¼ fs ¼ kx
y1 ¼ x; (f)

which, by Eq. (b), yields the output equations as follows

y1 ¼ x2 ; y2 ¼ kx1 (g)
 3.7 State-Space Representations 189

Example 3.7.2
For the cart–mass system in Figure 3.2.15, develop a state-space model, with the external force f as
the input and the displacement x2 of mass m2 and the force of spring k2 as the outputs. Also, for
this linear system, obtain the state and output equations in matrix form.
Solution
From Example 3.2.5, the coupled equations of motion are

m1 x€1 þ Bx_ 1 þ ðk1 þ k2 Þx1  Bx_ 2  k2 x2 ¼ f

(a)
m2 x€2 þ Bx_ 2 þ k2 x2  Bx_ 1  k2 x1 ¼ 0

with each being a second-order differential equation. Because the sum of the orders of the
equations is 2 þ 2 ¼ 4, four state variables are chosen as follows

z1 ¼ x1 ; z2 ¼ x_ 1 ; z3 ¼ x2 ; z4 ¼ x_ 2 (b)

where the notation zk is used for the state variables to avoid confusion in symbols. Differentiating
z1 and z3 immediately gives two state equations z_ 1 ¼ z2 ; z_ 3 ¼ z4 , where the definition of z2 and z4 in
Eq. (b) has been used. Now, differentiate z2 and z4 and substitute the results into the equations (a)
of motion, to obtain
m1 z_ 2 þ Bz2 þ ðk1 þ k2 Þz1  Bz4  k2 z3 ¼ f
(c)
m2 z_ 4 þ Bz4 þ k2 z3  Bz2  k2 z1 ¼ 0

It follows that the state equations of the system are

z_ 1 ¼ z2
1
z_ 2 ¼ ½ðk1 þ k2 Þz1  Bz2 þ k2 z3 þ Bz4 þ f 
m1
(d)
z_ 3 ¼ z4
1
z_ 4 ¼ ½k2 z1 þ Bz2  k2 z3  Bz4 
m2

The outputs are

y1 ¼ x2 ; y2 ¼ fs ¼ k2 ðx2  x1 Þ (e)

which gives the output equations as follows:

y1 ¼ z3
(g)
y2 ¼ k2 ðz3  z1 Þ

The state equations (d) can be cast in matrix form, x_ ¼ A þ Bu, with
190 3 Mechanical Systems

2 3
0 1 0 0 2 3
0 1 6 7 0
6 k þk 7
z1 6 1 2 B k2 B 7 6 7
Bz2 C 6  7 6 7
6 m m1 m1 m1 7; B ¼ 6 1 7; u ¼ f
x¼B C
@z3 A; A ¼ 6
1
7 6 7 (h)
6 7 6m1 7
6 0 0 0 1 7 4 5
z4 6 7 0
4 k2 5
B  k2  B 0
m2 m2 m2 m2

Also, the output equations can be written in the matrix form, y ¼ Cx þ Du, with

y1 0 0 1 0 0
y¼ ; C¼ ; D¼ (i)
y2 k2 0 k2 0 0

Example 3.7.3
Consider the motor–propeller assembly in Figure 3.3.10, which experiences a nonlinear drag
torque. Establish a state-space representation, with the motor torque as the input, and the
propeller rotation speed and the torque of the spring connecting the motor and propeller as the
outputs.
Solution
The equations of motion of the system, from Example 3.3.1, are given by

Jm θ€m þ bθ_ m þ kT θm  kT θp ¼ τm
(a)
Jp θ€p þ bD θ_ p sgnðθ_ p Þ þ kT θp  kT θm ¼ 0
2

According to these equations, four state variables are selected as follows

x1 ¼ θ m ; x2 ¼ θ_ m ; x3 ¼ θp ; x4 ¼ θ_ p (b)

Equation (b) automatically gives two state equations: x_ 1 ¼ x2 ; x_ 3 ¼ x4 . The other two state
equations are derived by substituting Eq. (b) into Eq. (a), which leads to

Jm x_ 2 þ bx2 þ kT x1  kT x3 ¼ τm
(c)
Jp x_ 4 þ bD x24 sgnðx4 Þ þ kT x3  kT x1 ¼ 0

It follows that the four state equations of the motor–propeller system are
 3.8 Dynamic Responses via MATLAB 191

x_ 1 ¼ x2
1
x_ 2 ¼ ðkT x1  bx2 þ kT x3 þ τm Þ
Jm
(d)
x_ 3 ¼ x4
1 
x_ 4 ¼ kT x1  kT x3  bD x24 sgnðx4 Þ
Jp

Note that the fourth state equation

 is nonlinear due to the drag torque. The system outputs are
y1 ¼ θ_ p and y2 ¼ τs ¼ kT θm  θp , which by Eq. (b) give the output equations as follows:

y1 ¼ x 4
(e)
y2 ¼ k T ð x1  x3 Þ

3.8 Dynamic Responses via MATLAB

So far, this chapter has been focused on the derivation of mathematical models for
mechanical systems and the corresponding transfer function formulations and state-
space representations. In this section, the simulation of dynamic responses of mechan-
ical systems by the software MATLAB is presented. The purpose here is twofold: to
show the utility of transfer function formulations and state-space representations in the
simulation and analysis of dynamic systems, and to investigate the dynamic character-
istics of mechanical systems in time-response calculations. One may refer to Appendix
B for a quick tutorial on MATLAB.

Example 3.8.1
Consider the spring–mass–damper system:

x ðtÞ þ cx_ ðtÞ þ kxðtÞ ¼ f ðtÞ

m€

with its parameters given by m = 20 kg, c = 80 N∙s/m, and k = 750 N/m. Assume a constant
external force (step input), f ðtÞ ¼ f0 = 15 N, and zero initial disturbances, xð0Þ ¼ 0 and x_ ð0Þ ¼ 0.
Plot the displacement and the resultant of the internal forces (the sum of the spring and damping
forces) of the system, for 0 ≤ t ≤ 3 s.
Solution
As shown in Example 3.2.8, the s-domain displacement of the system is given by

1
X ðsÞ ¼ FðsÞ (a)
ms2 þ cs þ k
192 3 Mechanical Systems

The resultant of the internal forces is fR ¼ cx_ þ kx, which in the s-domain is

FR ðsÞ ¼ ðcs þ kÞX ðsÞ (b)

It follows from Eqs. (a) and (b) that

1 f0
X ðsÞ ¼ (c)
ms2 þ cs þ k s

cs þ k f0
FR ðsÞ ¼ (d)
ms2 þ cs þ k s

where F ðsÞ ¼ f0 =s has been used.

With Eqs. (c) and (d), the time histories of xðtÞ and fR ðtÞ are computed by the following MATAB
commands in a script M-file:
m = 20; c = 80; k = 750;
f0 = 15; t_final = 3;
figure (1) % Displacement plot
syst1 = tf([f0],[m c k])
step(syst1, t_final, ‘k’)
xlabel(‘t’), ylabel(‘x (m)’), title(' ')
figure (2) % Resultant force plot
syst2 = tf(f0*[c k], [m c k])
step(syst2, t_final, ‘k’)
xlabel(’t’), ylabel(‘f_R (N)’), title(' ')
This produces Figure 3.8.1. Here, tf and step are MATLAB functions; see Tables B6 and B8 in
Appendix B.
As seen from Figure 3.8.1(a), the displacement x of the mechanical system under a step input is
oscillatory at the beginning, and it eventually settles to a constant value, which is known as the
steady-state response. This value is the static deflection of the spring at equilibrium; namely,
f0 15
xss ¼ ¼ ¼ 0:02 m. Likewise, the resultant force fR ðtÞ of the spring and damping forces
k 750
eventually settles at the value of kxss ¼ f0 ¼ 15 N, as shown in Figure 3.8.1(b).
 0.03 25

0.025
20

0.02
15

fR (N)
x (m)

0.015

10
0.01

5
0.005

0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
t (seconds) t (seconds)
(a) (b)

Figure 3.8.1 Dynamic response of the mechanical system in Example 3.8.1: (a) displacement x versus time; (b) resultant force fR versus
time
194 3 Mechanical Systems

Example 3.8.2
For the coupled motor–propeller system shown in Figure 3.3.10 of Example 3.3.1, let the param-
eter values be

Jm ¼ 1:2 kgm2 ; b ¼ 0:5 Nms; kT ¼ 5000 Nm; Jp ¼ 1:1 kgm2 ; bD ¼ 0:02Nms2

Consider a motor torque τm = 12 N∙m. Assume that the system is initially at rest. By a state-space
representation, compute and plot the propeller rotation speed θp _ and the spring torque
 
τs ¼ kT θm  θp of the coupled system, for 0 ≤ t ≤ 30 s. Also, investigate the effects of the torsional
spring and air damping on the system response.
Solution
To determine the system response, use the state and output equations obtained in Example 3.7.3
and the fixed-step Runge–Kutta method of order four given in Section 2.8. To this end, the
following commands contained in a script M-file:
global Jm b_r Kt Jp % System parameters
global b_D % Air damping torque
global Tm % Motor torque
% Input data
Jm = 1.2; b_r = 0.5; Kt = 5000; Jp = 1.1;
b_D = 0.02;
Tm = 12;
% Compute system response
t_stop = 30;
npts = 10001;
h = t_stop/(npts-1);
z0 = [0; 0; 0; 0]; % initial state
t = linspace(0, t_stop, npts);
z = zeros(4, npts);
z(:,1) = z0;
for i = 1: npts-1
tt = t(i);
zz = z(:,i);
f1 = RKfunc362(tt, zz);
f2 = RKfunc362(tt+h/2, zz+h/2*f1);
f3 = RKfunc362(tt+h/2, zz+h/2*f2);
f4 = RKfunc362(tt+h, zz+h*f3);
z(:,i+1) = zz +h/6*(f1+2*f2+2*f3+f4);
end

% Plot system response

figure(1)
theta_p = z(4,:); % Propeller rotation speed
plot(t,theta_p,‘k’)
grid
xlabel(‘Time, t (seconds)’)
 3.8 Dynamic Responses via MATLAB 195

ylabel(‘Speed, \theta_p (rad/s)’)

figure(2)
Ts = Kt*(z(1,:) – z(3,:)); % Spring torque
plot(t,Ts,’k’)
grid
xlabel(‘Time, t (seconds)’)
ylabel(‘Spring Torque, \tau_s (N-m)’)
and the following function M-file:
function f = RKfunc362(t,z)
global Jm b_r Kt Jp % System parameters
global b_D % Air damping torque
global Tm % Motor torque

f = zeros(4,1);
f(1) = z(2);
f(2) = (-Kt*z(1) – b_r*z(2) + Kt*z(3) + Tm)/Jm;
f(3) = z(4);
f(4) = (Kt*z(1) – Kt*z(3) – b_D*z(4)^2*sign(z(4)))/Jp;
are created. This plots the propeller speed and the spring torque against time in Figure 3.8.2.
As seen from Figure 3.8.2, the steady-state values of the propeller speed and spring torque are
about 15 rad/s and 4.5 N∙m, respectively. Due to the elastic coupling between the motor shaft ðJm Þ
and propeller shaft ðJp Þ, high-frequency oscillations appear in the system response; see Figures
3.8.3(a) and 3.8.3(b), which are obtained by zooming in on Figures 3.8.2(a) and 3.8.2(b).
To see the effect of air damping, set bD ¼ 0 and execute the above-mentioned M-files. This gives
the propeller speed plot given in Figure 3.8.4, which has a steady-state value of 24 rad/s.
Comparison of Figures 3.8.3(a) and 3.8.4 shows that air damping significantly reduces the
steady-state speed of the propeller (from 24 rad/s to 15 rad/s). Note that the system without air
damping is a linear system. From Example 3.3.7, the transfer function

sΘp ðsÞ kT
GðsÞ ¼ ¼  
Τ m ðsÞ ðJm s þ bÞ Jp s2 þ kT þ Jp kT s

can be used to compute the system response. For instance, the MATLAB commands
d = [Jm*Jp b_r*Jp (Jm+Jp)*Kt b_r*Kt];
n = Kt*Tm;
sys = tf(n,d);
Tspan = linspace(0, t_stop, npts);
step(sys, Tspan)
yield the same speed plot as shown in Figure 3.8.4.
 16

14

12
Speed, ωp (rad/s)

10

0
0 5 10 15 20 25 30
Time, t (seconds)
(a) (b)

Figure 3.8.2 Dynamic response of the coupled system in Example 3.8.2: (a) the propeller speed, ωp ¼ θ_ p ; and (b) the spring torque,
 
τs ¼ kT θm  θp
 3.8 Dynamic Responses via MATLAB 197

Speed, ωp (rad/s)
4

0
0.2 0.4 0.6 0.8 1. 1.2 1.4 1.6
Time, t (seconds)

(a)
12

10
Spring Torque, τs (N.m)

0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

Time, t (seconds)
(b)

Figure 3.8.3 Zooming in on the response of the coupled system in Example 3.8.2: (a) the propeller
speed; and (b) the spring torque

No Air Damping, bD = 0
25

20
Speed, ωp (rad/s)

15

10

0
0 5 10 15 20 25 30
Time, t (seconds)

Figure 3.8.4 Propeller speed of the coupled system without air damping
198 3 Mechanical Systems

Besides MATLAB, Simulink, which makes use of block diagram representation, is

useful for the modeling and simulation of mechanical systems. Usage of Simulink is
addressed in Section 6.6 and a tutorial on Simulink is given in Appendix B.2.

3.9 Dynamic Responses via Mathematica

This section illustrates the use of Wolfram Mathematica for the simulation of dynamic
responses of mechanical systems. It presents the implementation of the mathematical models
of the dynamic systems presented in Examples 3.8.1 and 3.8.2 in Wolfram Language. One
may refer to Appendix C for a tutorial on Wolfram Mathematica.

Example 3.9.1
Repeat Example 3.8.1 by using Mathematica.
Solution
From Example 3.8.1, the s-domain displacement and resultant force are given by

1
X ðsÞ ¼ FðsÞ (a)
ms2 þ cs þ k
FR ðsÞ ¼ ðcs þ kÞX ðsÞ (b)

Simulation of the system response takes the following two steps:

(i) implement the expressions in Eqs. (a) and (b) by using the built-in function
TransferFunctionModel; and
(ii) generate a system response by using the built-in function OutputResponse.

The code for the previous steps is shown in Figure 3.9.1.

Note the use of a replacement operator instead of a numerical value assignment. This is done in
order to be able to print the generated transfer functions in symbolic form and verify them with the
hand-derived results.
Since the governing second-order ordinary differential equation is linear with constant coeffi-
cients, a closed-form solution is expected to be generated by OutputResponse. This solution can
be shown by using the Print statement, as seen in Figure 3.9.2.
Plotting the generated responses is straightforward with the built-in function Plot. Formatting
the produced graphics is done by using the available options of the Plot function; see Figure
3.9.3. Refer to Wolfram Mathematica documentation for a detailed description of the shown
functions and plot-modifying options.
As expected, the computed values correspond to the depiction of responses on the graphs and
they are the same as obtained with MATLAB in Example 3.8.1.
 3.9 Dynamic Responses via Mathematica 199

Figure 3.9.1 Wolfram Language code for creating the required transfer functions and generating the
required responses

Figure 3.9.2 Printing implemented transfer functions and responses: (a) code; (b) output
Figure 3.9.3 Generating the required plots: (a) code for the plot of displacement vs. time; (b) plot of
the displacement vs. time; (c) code for the plot of resultant of internal forces vs. time; (d) plot of
resultant of internal forces vs. time
 3.9 Dynamic Responses via Mathematica 201

Example 3.9.2
Repeat Example 3.8.2 by Mathematica.
Solution
The coupled motor–propeller system is implemented by using the built-in Mathematica function
NonlinearStateSpaceModel. To this end, a nonlinear state-space model of the coupled
system, which is described by Eqs. (b), (d) and (e) in Example 3.7.3, is used. Implementation of
the state-space model in Wolfram Language is shown in Figure 3.9.4.
The specified responses: the propeller rotation speed and the spring torque – are computed by
using the OutputResponse function, and graphs are generated and formatted to the desired
appearance by using Plot with its options. The code and its output are shown in Figure 3.9.5.
If desired, we can zoom in on the generated graphs to observe the high-frequency oscillations due
to the elastic coupling between the motor shaft and the propeller shaft, as shown in Figure 3.9.6.
To see the effect of air damping, generate a response from the implemented state-space model
setting bD ¼ 0. Since the model was created in a symbolic form, with assignment of values to
variables that are not expected to change, we can use the replacement syntax to compute the
response for a system with no nonlinear air damping. Plotted in Figure 3.9.7 is the propeller speed
of the coupled system with and without air damping. Here, the words “linear system” refer to the
coupled system without damping, which is seen from Eq. (a) in Example 3.7.3, and the words
“nonlinear system” refer to the coupled system with air damping.

Figure 3.9.4 Implementation of the nonlinear state-space model developed in Example 3.7.3: (a) code;
(b) model matrix
Figure 3.9.5 Dynamic response of the coupled motor–propeller system: (a) code; (b) the propeller
speed; and (c) the spring torque
 3.9 Dynamic Responses via Mathematica 203

Figure 3.9.6 Zooming in on the response of the coupled motor–propeller system: (a) the propeller
speed; (b) the spring torque

Figure 3.9.7 Comparison of the propeller speed of the coupled system: (a) nonlinear system – the
coupled system with air damping; and (b) linear system – the coupled without air damping
204 3 Mechanical Systems

CHAPTER SUMMARY
This chapter introduces a systematic method for the derivation of equations of motion for
general mechanical systems. In this modeling method, three important keys, namely
Newton’s laws, basic elements, and free-body diagrams, are emphasized. After a mathemat-
ical model is developed, it is converted to two forms of model representation: transfer-
function formulation and state-space representation. These model representations are useful
for simulation and analysis. In addition, an energy method with Lagrange’s equations is
introduced for modeling of mechanical systems. Finally, the utility of MATLAB and
Mathematica to simulate dynamic responses of mechanical systems is illustrated.
Upon completion of this chapter, you should be able to:

(1) Understand and use the three important keys in modeling of mechanical systems.
(2) Derive equations of motion for spring–mass–damper systems, translational and rota-
tional systems, rigid-body systems in plane motion, and coupled multi-body systems.
(3) Establish transfer function formulations and state-space representations for given equa-
tions of motion.
(4) Apply Lagrange’s equations to derive the equations of motion of multi-body systems.
(5) Compute the dynamic response of mechanical systems by MATLAB and Mathematica.

REFERENCES
1. F. P Beer, E. R. Johnston, Jr., P. Cornwell, and B. Self, Vector Mechanics for Engineers: Dynamics,
11th ed., McGraw-Hill Education, 2015.
2. J. L. Meriam, L. G. Kraige, and J. N. Bolton, Engineering Mechanics: Dynamics, 8th ed., Wiley, 2015.

PROBLEMS
Section 3.1 Fundamental Principles of Mechanical Systems
3.1 A truck of mass m moves up on a slope of angle α; see Figure P3.1, where x is the
displacement of the truck, f is a pulling force generated by the vehicle engine, and g is the
gravitational acceleration. Assume that f ðtÞ ¼ f0 eσt þ mg sin α, where f0 is a constant.
Let the initial displacement and velocity of the truck be x0 and v0 , respectively

g
x
m f

Figure P3.1
 Problems 205

(a) Derive the differential equation of motion of the car.

(b) Solve the governing equation of part (a) for the car displacement x.
(c) Consider the parameters of the car: m = 2000 kg, α = 5◦, f0 ¼ 1500 N, σ ¼ 0:3=s, and
v0 ¼ 3 m=s. Plot the velocity of the car for 0 ≤ t ≤ 20 s.
3.2 In Figure P3.2, under the influence of gravity, a ball of mass m slides down on a smooth
slope (H > h), starting at point A with zero velocity. After leaving point B of the slope,
the ball is in a free fall, and it eventually lands on the floor at point C.

m
g
A B

H
h
D E C

L l

Figure P3.2

(a) Write down the kinetic energy and potential energy of the ball at points A and B.
You may take point D as a reference point for potential energy.
(b) For H = 2 m, h = 1.5 m and L = 3 m, determine the horizontal and vertical
components of the ball velocity at point C, and compute the distance l between
point E and point C.
3.3 In Figure P3.3, a soccer ball of mass m at point O is kicked with an angle θ. The kicking
force, which acts at the ball in a very short period of time, can be described by
f ðtÞ ¼ I0 δðtÞ, where δðtÞ is the Dirac delta function, and I0 is the impulse magnitude.
Consider the coordinate system Oxy, where the ball is initially at the origin O. Point A
where the ball lands on the floor has the coordinates x ¼ xA and y = 0. Ignore aero-
dynamic effects.

g
y f (t)

θ
O m A
x

Figure P3.3
206 3 Mechanical Systems

(a) Determine the horizontal distance xA of the ball.

(b) Determine the angle θ such that the ball drops at the farthest point.
(c) Consider the parameter values: m = 0.4 kg, θ ¼ 43:5°, I0 ¼ 16:8 kg  m=s. Plot the
trajectory of the ball in the xy-plane and determine the distance xA .

Section 3.2 Translational Systems

3.4 Consider the mechanical system in Figure P3.4.

(a) Draw free-body diagrams for the mass and auxiliary plots for the spring and
damping elements.
(b) Derive the governing equations of motion for the system.

x
k
k1
f
M

Figure P3.4

3.5 Consider the mechanical system in Figure P3.5, where y is a specified displacement.
(a) Draw free-body diagrams for the mass elements and auxiliary plots for the spring
and damping elements.
(b) Derive the governing equations of motion for the system.

x1 x2
k
k1 y
f
m1 m2

Figure P3.5

3.6 Consider the mechanical system in Figure P3.6, where f and q are external forces.
(a) Draw free-body diagrams for the mass elements and auxiliary plots for the spring
and damping elements.
(b) Derive the governing equations of motion for the system.
 Problems 207

x1 x3
k1

f q
m1 x2 m3
m2
c k

Figure P3.6

3.7 Consider the mechanical system in Figure P3.7, where there exists dry friction ðμD Þ
between mass m1 and the slope.

x2
x1
g f
k m2
m1
μD

Figure P3.7

(a) Draw free-body diagrams for the mass elements and auxiliary plot for the spring
element.
(b) Derive the governing equations of motion for the system.

3.8 Consider the mechanical system in Figure P3.8, where viscous friction (B) exists between
m1 and m2 , and dry friction ðμD Þ exists between m3 and the ground.
(a) Draw free-body diagrams for the mass elements and auxiliary plots for the spring
and damping elements.
(b) Derive the governing equations of motion for the system.

x2 x1 g

k1
m2 x3
c
B k2 f
m1 m3
μD

Figure P3.8
208 3 Mechanical Systems

3.9 Consider the mechanical system in Figure P3.9, where f and p are external forces; viscous
friction (B) exists between m1 and m2 ; and a massless pulley changes the direction of the
tension force of an unstretchable rope, which suspends a weight (M).
x2 x1
Massless
f
k2 pulley
m2 k3
k1
B
m1
c
g

Rope
M
y

Figure P3.9
(a) Draw free-body diagrams for the mass elements and auxiliary plots for the spring
and damping elements.
(b) Derive the governing equations of motion for the system.

3.10 Consider a quarter-car model shown in Figure P3.10, where m1 , m2 ; and m3 represent the
masses of the wheel–tire–axle assembly, car body, and engine, respectively; springs k,
k1 , and k2 model the elasticity of the tire, suspension, and engine mounts, respectively;
dampers c1 and c2 describe the energy dissipation of the suspension and engine mounts,
respectively; and yo is an displacement input describing the condition of road surface.
(a) Draw free-body diagrams for the mass elements and auxiliary plots for the spring
and damping elements.
(b) Derive the governing equations of motion for the car.

y3
m3

c2 k2

y2
g
m2

c1 k1

y1
m1

k yo

Figure P3.10
 Problems 209

3.11 In Figure P3.11, a spring–mass–damper system is connected to a massless rod of length

L and longitudinal rigidity EA. By using Table 3.2.1, develop a mathematical model for
the system.

Rod x1 x2

EA k c
f
m1 m2

Figure P3.11

3.12 A massless cantilever beam of length L and bending stiffness EI carries a spring–mass–
damper subsystem at the tip; see Figure P3.12, where p is an external force. The combined
beam–mass system can be used as a simplified model of airplane wings carrying engines or
crane arms lifting loads. Consider gravity. Develop a mathematical model of the combined
system. Refer to Table 3.2.1 for the effective spring coefficient of the beam.

Beam

EI

c k

M
y

Figure P3.12

3.13 For the system in Figure P3.4, derive the transfer function with force f as the input and
velocity x_ as the output. Also, state the order of the transfer function.
3.14 For the system in Figure P3.5, derive the following two transfer functions:
(a) The transfer function G1 ðsÞ with displacement y as the input and velocity x_ 1 as the
output.
(b) The transfer function G2 ðsÞ with force f as the input and displacement x2 as the
output.
210 3 Mechanical Systems

Section 3.3 Rotational Systems

3.15 Power transmission belts are commonly seen in automobiles and machines. Shown
in Figure P3.15(a) is a belt–drive system, in which a looped belt of flexible
material links two pulleys. Driven by the drive pulley (connected to a motor),
the belt brings the driven pulley to rotate. The inertia of the belt, compared to
that of the pulleys, is negligible. Therefore, the belt can be viewed as two springs
that connect the pulleys; see Figure P3.15(b), where τ is an external torque applied
to the drive pulley.
(a) Draw free-body diagrams for the mass elements and auxiliary plots for the spring
elements.
(b) Derive the governing equations of motion for the belt–drive system.
(c) Determine the transfer function with τ as the input and θ_ 2 as the output. Also, state
the order of the transfer function.

θ1 R1 R2
I1 I2
O1 O2 θ2
τ
Drive Driven k
pulley Belt pulley
(a) (b)

Figure P3.15

3.16 A disk–shaft system is shown in Figure P3.16(a), where a torque τ is applied to

the left disk. A model of the system is shown in Figure P3.16(b), where the shafts
are modeled as torsional springs ðk1 ; k2 Þ; the mass moments of inertia of the disks are
denoted by I1 and I2 ; and double-headed arrows are used to describe the torque and
disk rotations.
(a) Draw free-body diagrams for the mass elements and auxiliary plots for the spring
elements.
(b) Derive the governing equations of motion for the disk–shaft system.
(c) Derive the transfer function with τ as the input and the internal torque of spring k2
as the output.
 Problems 211

Shafts
τ
τ k1 k2
I1 I2

θ1 θ2
Disks

Figure P3.16

3.17 In Figure P3.17, an L-shaped rigid bar is pivoted at point O and constrained by a spring
(k) at tip A and a damper (c) at tip B. The mass moment of inertia of the bar with respect
to O is IO . An external (horizontal) force f is applied to the bar at point C. Assume a
small rotation, jθj ≪ 1. In this problem, gravity is not considered.

k
A

C f (t)

h
a

θ
B
O

c
l

Figure P3.17

(a) Derive the linear differential equation governing the rotation θ of the bar.
ΘðsÞ
(b) Determine the transfer function .
F ðsÞ
3.18 In Figure P3.18, a mass (M) is linked to a level by a spring ðkl Þ, which is pinned at point
O, is subject to a horizontal force f, and constrained by a viscous damper (c). Denote
the mass moment of inertia of the level with respect to point O by IO . Consider a small
rotation of the level.
(a) Derive the linear differential equations of motion for the mass–level system.
X ðsÞ
(b) Determine the transfer function . Also, state the order of the dynamic system.
FðsÞ
212 3 Mechanical Systems

x
k kl
θ
M

a
O

b cl

c
f

Figure P3.18

3.19 A coupled disk–mass system is shown in Figure P3.19, where p is an external force
applied to the mass. In this problem, gravity is not considered.
(a) Derive the governing equations of motion for the system.
(b) Derive the transfer function, from force p to rotation θ.

R c
O

Io θ
k

m
y
p

Figure P3.19

3.20 In Figure P3.20, a cart of mass M carrying a single pendulum (mb , L) moves vertically.
Constrained by a spring and damper pair (k, c), the cart is subject to an external force f.
(a) Draw free-body diagrams for the mass elements.
(b) Derive the nonlinear equations of motion for the cart–pendulum system.
(c) With the results from part (b), obtain two independent differential equations of
motion about y and θ.
(d) Determine the expression of the tension force of the pendulum in terms of y and θ.
 Problems 213

c k

M
g
y

f
L
θ

mb

Figure P3.20

3.21 In Figure P3.21, two rotating shafts are coupled by a torsional spring (k). A torque τ is
applied to the left shaft and a torsional damper (c) is connected to the right shaft.
(a) Draw free-body diagrams for the mass elements and auxiliary plots for the spring
and damping elements.
(b) Derive the governing equations of motion for the system.
(c) Determine the transfer function with the torque as the input and the rotation speed
of the right shaft as the output.

J1 θ1 × k J2 θ2 × c

τ × ×
b1 b2

Figure P3.21

3.22 In Figure 3.22(a), a rotating shaft is coupled to a pulley by a torsional spring (k) and
the pulley suspends a mass (M) by an unstretchable rope. An external torque τ is
applied to the shaft. The side view of the pulley is shown in Figure P3.22(b). Consider
gravity.
(a) Derive the governing equations of motion for the combined shaft–pulley–mass
system.
(b) Determine the transfer function from the torque τ to the displacement y of the
mass.
214 3 Mechanical Systems

Pulley Pulley

J θ1 × k θ2
θ2 R

τ × O g
Io Io
b

y
y
M M

(a) (b)

Figure P3.22

3.23 Consider the geared system in Figure P3.23, where double-headed arrows are used to
describe torque τ and shaft rotations θ1 and θ2 .
(a) Draw free-body diagrams for the mass elements and auxiliary plots for the spring
and gear pair.
(b) Derive two independent equations of motion for the geared system.
(c) Determine the transfer function from torque τ to rotation speed θ_ 2 . Also, state the
order of the geared system.

J1 × n1
τ
× J2
k ×
×
θ1 ×
b1
n2 ×
θ2 b2

Figure P3.23

3.24 In Figure P3.24(a), two shafts are coupled by a compound gear train. The gear train
consists of two pairs of gears, as shown in Figure P3.24(b). The right shaft experiences a
drag torque ðbD Þ, which is described by Eq. (3.3.12).
(a) Draw free-body diagrams for the mass elements and auxiliary plots for the
gear train.
(b) Derive the nonlinear equation of motion for the geared system in terms of θ_ 2 .
 Problems 215

J1 × n1
τ
× n3
n2
× n1
θ1 × × J2
b1 ×
n2 × n3 n4
n4 ×
θ2 bD
b2
(a) (b)

Figure P3.24

Section 3.4 Rigid-Body Systems in Plane Motion

3.25 A rigid body is mounted on two springs at its lower corners; see Figure P3.25, where G
is the center of mass of the body, a force f is applied at the upper-right corner A of the
body, yG is the vertical displacement of the mass center, and θ is the rotation angle of
the body. Assume that the body moves in the vertical direction, and rotates with small
angle, jθj≪1. The springs are undeformed if both yG and θ are zero. Gravity is not
considered. Derive the equations of motion for the elastically supported rigid body, in
terms of yG and θ.

f
yG
A
m, IG
θ
G
b
a

k2
k1

Figure P3.25

 
3.26 In Figure P3.26, a mass (m) is suspended from a movable pulley mp ; IG that is
 
elastically supported by a spring kp . The rope suspending the mass m is unextend-
ible and it is connected to a spring and damper pair (k, c). The spring and damping
elements are undeformed when both yG and θ are zero. Consider gravity. Derive
two independent equations of motion for the coupled pulley–mass system about yG
and θ.
216 3 Mechanical Systems

kp g

θ
G R
yG
mp, IG

m
c k
z

Figure P3.26

3.27 Consider the pulley–mass system in Figure P3.27, where two masses ðm1 ; m2 Þ are
suspended from a movable pulley ðmp ; IG Þ that is elastically supported by a spring
ðkp Þ. The masses are also constrained by a spring and damper pair (k, c). The spring and
damping elements are undeformed when both yG and θ are zero. Consider gravity.
Derive two independent equations of motion for the coupled pulley–mass system about
yG and θ.

kp g

θ
G R
yG
mp, IG

m1 m2
z1 z2

k c

Figure P3.27
 Problems 217

3.28 A cylinder rolls on a slope without slipping, as shown in Figure P3.28, where f is an
external force, G is the mass center of the cylinder, A is the cylinder–slope contact point,
and x is the displacement of the mass center. The spring is undeformed when x is zero.
Let m and IG be the mass and mass moment of inertia with respect to G, respectively.
Derive the equation of motion of the system.

g
f

R
k θ
G

A
α

Figure P3.28

3.29 In Figure P3.29, a cart carrying a cylinder is pulled by a force f. The cylinder rolls on the
cart without slipping and it is constrained by a spring and damper pair (k, c). Denote
the mass and mass moment of inertia with respect to the mass center G of the body by m
and IG , respectively. The spring and damper are undeformed when the rotation angle θ
of the cylinder is zero. Derive the equations of motion of the system in terms of x and θ.

x
θ

k R c

G
f
mc

Figure P3.29

3.30 A nonuniform rigid bar is suspended by two strings of equal length l; see Figure
P3.30, where a horizontal force f is applied to the bar at point A; the string
attachment points A and B and the mass center G of the bar are on a straight
0
line. The distance between points O and O is the same as the length of the bar;
namely, OO0 ¼ a þ b. With the geometric configuration, the bar in motion does
not have rotation. Denote the mass and mass moment of inertia with respect to
218 3 Mechanical Systems

the mass center G of the body by m and IG , respectively. Derive the equation of
motion of the bar in θ. Also, determine the tension forces of the strings.

x
g
O O′
y l l
θ

f
A G B

a b

Figure P3.30

3.31 In Figure P3.31, a uniform rigid bar is pinned at point O of a massless cart, which is
elastically supported by a spring (k). The cart can only move vertically, and it is subject
to an external force f. The spring is undeformed when the cart displacement y is zero.
Derive the equations of motion for the system.

k
Massless
cart
O

y
g f

θ m, l

Figure P3.31

3.32 A uniform rigid bar of mass m and length l is suspended by a string of length a;
see Figure P3.32, where Oxy is a Cartesian coordinate system, and G is the mass
center of the bar. Derive the equations of motion for the string–bar system in
terms of β and θ.
 Problems 219

x
O
y

a
β

A
g
m, l

θ
G

Figure P3.32

Section 3.5 Energy Approach

3.33 Consider the mechanical system in Figure P3.9.
(a) Write down the kinetic energy, potential energy, and Rayleigh dissipation function
of the system.
(b) Derive the equations of motion by Lagrange’s equations.
3.34 Consider the cart–pendulum system in Figure P3.20.
(a) Write down the kinetic energy, potential energy, and Rayleigh dissipation function
of the system.
(b) Derive the equations of motion by Lagrange’s equations.
3.35 Consider the elastically supported rigid body in Figure P3.25, without gravity.
(a) Write down the kinetic energy, potential energy, and Rayleigh dissipation function
of the system.
(b) Derive the equations of motion by Lagrange’s equations.
3.36 Consider the pulley–mass system in Figure P3.27.
(a) Write down the kinetic energy and potential energy of the system.
(b) Derive the equations of motion by Lagrange’s equations.
3.37 Consider the string-suspended rigid body in Figure P3.30.
(a) Write down the kinetic energy and potential energy of the system.
(b) Derive the equations of motion by Lagrange’s equations.
220 3 Mechanical Systems

Section 3.6 Block Diagrams of Mechanical Systems

3.38 Draw a block diagram for the spring–mass–damper system Figure P3.5, with f and y as
the inputs, x1 and x2 as the outputs.
3.39 Consider the mechanical system in Figure P3.9. By ignoring gravity, draw a block
diagram with p as the input, and y and the spring force of k2 as the outputs.
3.40 For the belt–drive system in Figure P3.15, draw a block diagram, with the toque τ as the
input, and the rotation speed θ_ 2 and the force of the upper spring as the outputs.
3.41 For the disk–mass system in Figure P3.19, draw a block diagram, with the force p as the
input, and the rotation speed θ_ and the displacement y as the outputs.
3.42 For the geared system in Figure P3.23, draw a block diagram, with the torque τ as the
input, and the rotation speed θ_ 2 and the torque of the spring k as the outputs.
3.43 For the elastically supported rigid body in Figure P3.25, draw a block diagram, with
the force f as the input, and the internal forces of the springs ðk1 ; k2 Þ as the outputs
3.44 For the combined cart–cylinder system in Figure P3.29, draw a block diagram, with
force f as the input, and the rotation speed θ_ and the spring force as the outputs.
Section 3.7 State-Space Representations
3.45 For the system in Figure P3.5, develop a state-space model (state equations and output
equations), with f and y being the inputs and x1 and x2 the outputs. Also, cast the state
and output equations in matrix form.
3.46 Consider the system in Figure P3.6. Let forces f and q be inputs. Let the displacements
of the masses m1 and m2 and the internal force of the damper c be outputs. Obtain a
state-space representation (state and output equations) for the system. Also, cast the
state and output equations in matrix form.
3.47 For the quarter-car model in Figure P3.10, develop a state-space model, with yo as the
input, the engine displacement ðy3 Þ and the resultant force of the engine mounts as the
outputs. For this problem, gravity is ignored. Also, cast the state and output equations
in matrix form.
3.48 For the belt–drive system in Figure P3.15, develop a state-space model, with torque τ as
the input, and rotation speed θ_ 2 and the internal force of the upper spring as the
outputs. Also, cast the state and output equations in matrix form.
3.49 Consider the coupled disk–mass system in Figure P3.19. Ignore gravity. Develop a
state-space model, with force p as the input, and the rotation θ of the disk and the
internal force of the spring connecting the mass and disk as the outputs. Also, cast the
state and output equations in matrix form.
3.50 For the cart–pendulum system in Figure P3.20, develop a state-space model, with force
f as the input, and displacement y and rotation θ as the outputs. Note that this is a
nonlinear system.
3.51 Consider the combined shaft–pulley–mass system in Figure P3.22. Let the torque τ and
 Problems 221

gravitational force Mg be the inputs. Let the displacement of the mass and tension force
of the rope be the outputs. Develop a state-space model for the system. Also, cast the
state and output equations in matrix form.
3.52 Consider the elastically supported rigid body in Figure P3.25. Assume a small rotation
of the body. Develop a state-space model for the system, with f as the input and the
spring forces as the outputs. Also, cast the state and output equations in matrix form.
Sections 3.8 and 3.9 Simulation of Dynamic Response via Software
3.53 Consider the spring–mass–damper system in Figure P3.5, with the parameter values
given by m1 ¼ 20 kg, m2 ¼ 30 kg, k ¼ 2000 N=m, k1 ¼ 3000 N=m, c ¼ 75 N  s=m.
Assume zero initial conditions.
(a) Assume that f ¼ 0 and y = 0.5 m (step input). Plot the displacement x2 versus time
by MATLAB.
(b) Assume that f ¼ 45δðtÞ N·s (impulse input) and y ¼ 0, where δðtÞ is the Dirac delta
function. Note that the force of the spring k1 is zero in this case. Plot the displace-
ment x2 versus time by MATLAB.
Hint: In parts (a) and (b), you can obtain the corresponding transfer functions first.
3.54 Consider the belt–drive system in Figure P3.15, with the following non-dimensional
parameters as follows: I1 ¼ 20, I2 ¼ 35, k ¼ 600, R1 ¼ 1 and R2 ¼ 1:5. Assume zero
initial conditions. Consider a step input τ ¼ 40. Use MATLAB to plot the following
two outputs versus time: (a) the rotation speed θ_ 2 of the driven pulley; and (b) the force
of the upper spring. Hint: You can derive the corresponding transfer functions first.
3.55 Consider the elastically supported rigid body in Figure P3.25, with the assigned param-
eter values: m ¼ 50 kg, IG ¼ 32 kg  m2 , k1 ¼ 1200 N=m, k2 ¼ 1600 N=m, a = 1.2 m, b =
0.8 m. Assume zero initial conditions. Consider an impulse force f ¼ 10δðtÞN  s. Use
MATLAB to plot the time response of the system: yG ðtÞ and θðtÞ, for 0 ≤ t ≤ 10 s.
3.56 Consider the string-suspended rigid bar in Figure P3.30, with the parameters assigned
as follows: m ¼ 40 kg, IG ¼ 18 kg  m2 , l = 0.6 m, a ¼ 0:6 m, b ¼ 0:4 m. The initial
conditions of the system are given by θð0Þ ¼ π=9 rad and θ_ ð0Þ ¼ 0.
(a) Obtain a state-space model for this nonlinear system.
(b) Let the applied force f at point A be f ¼ 20 sinð3tÞ N. With the model of part (a),
plot the response θðtÞ in degree versus time ð0 ≤ t ≤ 10 sÞ by the MATLAB function
ode45.
(c) Plot θðtÞ in degree versus time ð0 ≤ t ≤ 10 sÞ for f ¼ 20 sinð16tÞ N and compare the
result with that of part (b).
3.57 Solve Problem 3.53 by Mathematica.
3.58 Solve Problem 3.54 by Mathematica.
3.59 Solve Problem 3.55 by Mathematica.
3.60 Solve Problem 3.56 by Mathematica.
4 Electrical Systems

Contents
4.1 Fundamentals of Electrical Systems 223
4.2 Concept of Impedance 230
4.3 Kirchhoff’s Laws 242
4.4 Passive-Circuit Analysis 244
4.5 State-Space Representations and Block Diagrams 258
4.6 Passive Filters 271
4.7 Active-Circuit Analysis 277
4.8 Dynamic Responses via Mathematica 297
Chapter Summary 306
References 306
Problems 307

Practically every modern engineered dynamic system has electrical components such as
motors, sensors, controllers, or, at the very least, power sources. Therefore, an understand-
ing of the physical processes occurring in the typical electrical circuits, and the ability to
model behavior of an electrical subsystem are essential for anyone interested in dynamic
systems.
Similarly to the treatment of mechanical systems, the three keys in modeling of electrical
systems are identified as follows:

• fundamental principles, which are Kirchhoff’s current and voltage laws, as presented in
Section 4.3;
• basic elements, which include resistors, inductors, and capacitors, as described in Sections
4.1 and 4.2; and
• two ways of analysis, which are the loop method and the node method, as presented in
Section 4.4.

222
 4.1 Fundamentals of Electrical Systems 223

By a chosen way of analysis (either the loop method or the node method), Kirchhoff’s laws
are applied to the basic elements, resulting in a mathematical model for the electrical system
in consideration, which is a mixture of a set of differential and algebraic equations. This
three-key modeling concept is illustrated throughout this chapter.
A slightly different way of analysis, based on the described three-key modeling concept, is
presented in Section 4.5, where alternative models (state-space representations and block
diagrams) of the electrical circuits are derived.
Furthermore, in Sections 4.6 and 4.7, Kirchhoff’s laws and derived models of basic
elements are used to derive mathematical models of passive filters as well as active electrical
circuits and glean an insight into their operations. Numerical analysis of electrical circuits
behavior is presented in Section 4.8.

4.1 Fundamentals of Electrical Systems

Electrical and electronic components constitute important building blocks of many dynamic
systems. The most commonly encountered examples of electrical subsystems include power
supplies, controllers, sensors, and motors. Understanding their time-dependent behavior
becomes crucial for gaining an ability to model, analyze, design, and control a variety of
dynamic systems.

4.1.1 Definitions
An electrical system is often called an electrical circuit. A circuit consists of interconnected
circuit elements such as, for example, resistors, capacitors, inductors, or sources of energy.
There exist two main groups of circuit elements: active and passive.
Passive elements, usually called loads, dissipate or temporarily store the energy supplied to
the electrical circuit. They are represented by resistors, inductors, and capacitors.
Active elements are the sources that supply an electrical circuit with energy. Any type of
the energy-generating device such as chemical battery, solar/photovoltaic cells assembly,
thermocouple junction, or mechanical generator can serve as an active element in the
electrical circuit. Conversion of some kind of energy, for example, chemical for a battery,
thermal for a thermocouple, or mechanical for a generator, into electrical charge is the
common trait of all sources.
The active elements are commonly modeled as ideal sources of either current or voltage.
An ideal current source always supplies the specified current regardless of the voltage
required by the circuit, while an ideal voltage source always provides the specified
voltage regardless of the current generated in the circuit. A battery is considered an ideal
voltage source even though its performance is influenced by the heat generated as current
flows through the load circuit. If an impact of the produced heat cannot be ignored, the
battery can be modeled as an ideal voltage source connected to an internal resistor.
224 4 Electrical Systems

An electrical circuit generally includes one or more active elements. The energy provided
by a source is considered an input to the electrical system.
The electrical circuit theory uses current and voltage as primary variables to describe the
behavior of any electrical circuit.
Physically, current is the flow of electrons inside the circuit. Mathematically, it is expressed
as the time rate of change of electrical charge through a specific area:

dQ
i¼ (4.1.1)
dt

where i is the current, t denotes time, and Q is the electrical charge, or number of electrons
passing through the cross-section of a wire. Current is measured in amperes (A), and charge is
measured in coulombs (C). A negative one coulomb is equivalent to the electrical charge in
6:2415  1018 electrons. Similarly, a positive one coulomb is equivalent to the electrical
charge in 6:2415  1018 protons. It can also be defined as the amount of electrical charge
transported in one second by a constant current of one ampere.
The charge, expressed in terms of current and time, is
ð
QðtÞ ¼ idt (4.1.2)

The positive direction of the current flow is defined as opposite to the actual movement of
electrons inside the circuit. Hence, it can be visualized as a flow of positively charged
particles.
Voltage is defined as the work needed to move a unit charge between two points in a
circuit:

dW
v¼ (4.1.3)
dQ

where v is the voltage, Q is the electrical charge, and W denotes the work. One joule of work
performed moving one coulomb of charge along the electrical circuit yields the unit voltage
called the volt (V).
The voltage between any two points in a circuit can be also defined as the difference
between electric potentials of these points as shown in Figure 4.1.1, where e1 and e2 denote
electric potentials at the entrance and the exit
terminals of a circuit element respectively, i is e 1 Circuit
e2
the current flowing across this element, and v is element
i
the voltage computed as
v
v ¼ e 1  e2 (4.1.4)
Figure 4.1.1 Electrical circuit
terminology
 4.1 Fundamentals of Electrical Systems 225

The voltage at any point of an electrical circuit is represented as a difference between

electric potentials of this point and some selected reference point. The ground, or point of
zero electric potential, is commonly chosen as that reference point.
By convention, the element’s high-voltage terminal is denoted by the “þ” sign, while the
low-voltage terminal is denoted by the “” sign.
The sign of the voltage difference indicates whether the energy was generated or consumed
by the circuit element. A negative voltage difference across an element corresponds to energy
generation, while a positive difference indicates
energy consumption. To illustrate that concept con- e1 + i b1
sider the direct current (DC) electrical circuit +
depicted in Figure 4.1.2. Passive
circuit ve Battery
The electrons traveling across this circuit are vb
element
repelled by the negative terminal of the battery
b2 –
and attracted by its positive one. Thus, their motion e2 –
i
is clockwise, and, by definition, the positive flow of
the current becomes counterclockwise (as shown by
the arrows). Let vb and ve be the respective voltage
Figure 4.1.2 DC electrical circuit
differences across the battery and across the passive
circuit element. By definition, vb ¼ b2  b1 and
ve ¼ e1  e2 . The current gains energy inside the battery, flowing from the low (b2 ) to high
(b1 ) voltage terminal, so the potential difference across the battery is negative: vb < 0. On the
other hand, the current loses energy inside the passive circuit element, which makes
ve ¼ e1  e2 > 0, and e1 > e2 (as shown in Figure 4.1.2). Customarily ve is referred to as
the voltage drop, and vb – as the voltage gain.
According to the law of conservation of energy, in a closed circuit the energy supplied by
the battery must be fully consumed by the circuit elements. Hence, vb þ ve ¼ 0.
While the voltage drop and voltage gain have different signs as shown above, the opinion
on which one is considered positive differs from one literature source to another. In this
book, we adopt the convention that assigns a positive sign to voltage gain and a negative sign
to voltage drop.
In addition to voltage and current, electric circuits are characterized by power. The power
applied to a passive element, or generated by an active element, is a product of voltage across
that element and current flowing through it:

P ¼ vi (4.1.5)

This expression can be easily derived using the definition of power as work performed in a
unit time:

dW dW dQ dW dQ
P¼ ¼ ¼ ¼ vi (4.1.6)
dt dt dQ dQ dt
226 4 Electrical Systems

The unit of power is the watt (W), which corresponds to one joule of work performed in
one second.
Accordingly, energy consumed or generated by an element is

ðt
W ¼ Pdt (4.1.7)
0

4.1.2 Basic Elements

A set of symbols, used to indicate active and passive elements on the electric circuit diagrams,
is shown in Figure 4.1.3, where the symbols are as follows: (a) current source, (b) voltage
source, (c) battery, (d) resistor, (e) inductor, (f) capacitor, (g) ground, and (h) terminals.
Understanding the voltage–current relations of individual circuit elements is paramount
to the ability to model and analyze the behavior of any circuit.

Resistors
A resistor is a passive circuit element that dissipates energy. A simple light bulb is an example
of a resistor, which dissipates electrical energy by converting it into light and heat. The
behavior of a common resistor is governed by the empirically derived Ohm’s law, which
states that the electric current flowing through a resistor is directly proportional to the
voltage difference across that resistor:
v
i¼ (4.1.8)
R

where i is the current, v denotes voltage, and R is the resistance.

Figure 4.1.3 Basic elements of electrical circuits

 4.1 Fundamentals of Electrical Systems 227

Ohm’s law specifies the linear relation between voltage and current through a resistor, and
is often written in the form:

v ¼ iR (4.1.9)

If a resistor obeys Ohm’s aw, it is called linear resistor. The resistance of a linear resistor is a
constant, the value of which depends on the resistor’s material, geometry, and temperature.
For example, the resistance of a nonideal wire can be expressed as:

ρL
R¼ (4.1.10)
A

where L and A denote length and cross-sectional area of the wire respectively, and ρ is the
temperature-dependent resistivity of the wire material.
In this chapter all the resistors are considered linear, and all the connecting wires are
assumed to be perfect conductors and their resistance is neglected, i.e. the voltage drop
across an ideal wire is zero.
The unit of resistance is the ohm (Ω). A resistor with R ¼ 1 Ω experiences a voltage drop of
1 V when the current flowing through it equals 1A.
The power dissipated by a linear resistor is

v2
P ¼ vi ¼ ¼ i2 R (4.1.11)
R

Capacitors
A capacitor is a passive circuit element that
stores energy in the form of electrical charge.
The simplest capacitor circuit can be visu-
alized as a pair of parallel plates (a capacitor)
connected to a battery as shown in Figure
4.1.4. In this circuit the battery will transport
charge from one plate to the other until the Figure 4.1.4 A capacitor
voltage difference that results from the charge
build-up on the capacitor equals battery voltage vb .
A capacitor is characterized by its capacitance – a measure of how much charge can be
stored for a given voltage difference across the element. The mathematical expression of
capacitance is

Q
C¼ (4.1.12)
v
228 4 Electrical Systems

where Q is the stored charge, v denotes voltage, and C is the capacitance. The unit of
capacitance is the farad (F), which corresponds to a one coulomb of charge stored by the
capacitor experiencing a voltage drop of one volt.
For a parallel-plate capacitor, depicted in Figure 4.1.4, the capacitance depends on its
geometry and the material such as:

εA k ε0 A
C¼ ¼ (4.1.13)
d d

where A is the area of the flat parallel metallic plates of the capacitor, k is the
relative permittivity of the dielectric material between the plates (k ≈ 1 for the air),
ε0 ¼ 8:854x1012 F=m denotes the free-space permittivity, d is the distance between plates,
and ε ¼ ε0 k is the permittivity.
For a capacitor, the relation between a current through it and a voltage difference across it
is not linear. According to Eq. (4.1.12), the capacitor voltage is

Q
v¼ (4.1.14)
C

Substituting the expression for charge, defined by Eq. (4.1.2), we obtain

ð ðt
1 Q0 1
v¼ idt ¼ þ idt (4.1.15)
C C C
0

where Q0 is a charge on the capacitor at the time t ¼ 0. For simplicity of computations, it is

assumed that Q0 ¼ 0. Hence, the relation between capacitor voltage and current is expressed as:

ðt
1
Integral form v¼ idt
C (4.1.16)
0
dv
Derivative form i¼C
dt

The energy stored in a capacitor is derived by integrating the expression for power (Eq.
(4.1.5)) as stated in Eq. (4.1.7). Zero initial conditions are assumed:

ðt ðt ðt 
dv 1
W ¼ Pdt ¼ vidt ¼ v C dt ¼ Cv2 ðtÞ (4.1.17)
dt 2
0 0 0
Inductors
An inductor is a passive circuit element that stores energy in the form of a magnetic field.
The simplest inductor circuit can be visualized as a solenoid – a coil of wire wound into a
tight helix, which may or may not be wrapped around a ferromagnetic core (inductor),
connected to a source, as shown on Figure 4.1.5. An electric current is always accompanied
 4.1 Fundamentals of Electrical Systems 229

by a magnetic field – flux – around the conductor.

In the case of a solenoid conductor this flux is
concentrated almost entirely inside the solenoid,
while the outside field is weak. The longer the
solenoid, the more pronounced is this effect of a
nearly uniform flux inside the solenoid and negli-
gible one on the outside.
According to Faraday’s law (see Section 6.3.1 Figure 4.1.5 An inductor
for more detailed information), the relation
between flux and the current flowing through the inductor is

ϕ ¼ Li (4.1.18)

where ϕ is the flux, i denotes current, and L is the inductance. The inductance is a constant that
depends on solenoid geometry such as the length l, cross-sectional area of a coil A, and the
number of coils N, as well as on the material of the solenoid and its core, if one is present. The
mathematical expression for the induction of the solenoid, shown in Figure 4.1.5, is as follows:

μN 2 A
L¼ (4.1.19)
l

where μ denotes the magnetic permeability of the solenoid. The unit of inductance is the
henry (H).
The relation between flux and the voltage is expressed in integral form:
ð
ϕ ¼ vdt (4.1.20)

Equations (4.1.18) and (4.1.20) can be used to derive the expression for the voltage–
current relation of an inductor.

ðt
1
Integral form i¼ vdt
L (4.1.21)
0
di
Derivative form v¼L
dt

The energy stored in an inductor is derived using the Eqs. (4.1.5), (4.1.7), and the deriva-
tive form of Eq. (4.1.21). Zero initial conditions are assumed:

ðt ðt 
ðt
di 1
W ¼ Pdt ¼ vidt ¼ i L dt ¼ Li2 ðtÞ (4.1.22)
dt 2
0 0 0
230 4 Electrical Systems

Active Elements
Active circuit elements include energy sources described in Section 4.1.1, and electronic
devices such as integrated circuits, which include amplifiers, transistors, and logic gates.
While a detailed discussion on integrated circuits is beyond the scope of this book, amplifiers
will be presented later in this chapter (see Section 4.7).

4.2 Concept of Impedance

By consuming energy, passive circuit elements “resist” the
I(s) V(s)
flow of current. This opposition to a current is called imped- Z(s)
ance, and is represented by a voltage drop across the element.
Impedance is often viewed as a generalization of the resist- Figure 4.2.1 Impedance
ance concept, and as an analog of a transfer function that
takes current as an input and yields voltage as an output, as
shown in Figure 4.2.1. Impedance is typically expressed in the Laplace domain, and is
derived by taking the Laplace transform of the time domain expression of the element’s
voltage–current relationship,

4.2.1 Definition
Impedance is defined as the ratio of the voltage drop across the circuit element to the current
flowing across that element. Traditionally it is expressed in the Laplace domain as:

LfvðtÞg VðsÞ
Z¼ ¼ (4.2.1)
LfiðtÞg IðsÞ

The relations between current and voltage for the described earlier passive circuit elements
are used to derive their impedance expressions.
A resistor’s impedance is obtained by taking the Laplace transform of Eq. (4.1.9):

Lfvg ¼ VðsÞ ¼ Lfi Rg ¼ RIðsÞ (4.2.2)

Then
VðsÞ
ZRESISTOR ¼ ¼R (4.2.3)
IðsÞ

Using the derivative form of Eq. (4.1.16), the capacitor’s impedance is obtained:

dv
Lfig ¼ IðsÞ ¼ L C ¼ C s VðsÞ
dt
1
ZCAPACITOR ¼ (4.2.4)
Cs
 4.2 Concept of Impedance 231

For an inductor the derivative form of the Eq. (4.1.21) is used:


di
Lfvg ¼ VðsÞ ¼ L L ¼ L s IðsÞ
dt
ZINDUCTOR ¼ L s (4.2.5)

4.2.2 Combination of Impedances

A typical electrical circuit contains interconnected
+ I(s)
multiple passive and active elements. Knowledge of
the elements’ arrangement is important for the cor- Equivalent
Source impedance
rect modeling of the circuit since different ways of of of the
connecting the same elements yield differently behav- energy circuit
V(s) Zeq (s)
ing circuits.
Circuit analysis typically pursues two goals, the – I(s)
first one being the derivation of the voltage–current
relation for the circuit as an entity, and the second
one being the computation of the voltages across and Figure 4.2.2 Equivalent imped-
the currents through every component of the circuit. ance of a circuit
The former task is more easily fulfilled when the
circuit in question is transformed into an equivalent
form, as shown in Figure 4.2.2. Then, the computation of currents and voltage drops for
every passive circuit element is straightforward with the use of the previously presented
voltage–current relations for resistors, capacitors, and inductors.
To find the impedance of a circuit we first need to derive the rules for computing
equivalent impedance for two interconnected elements. Obviously, this equivalent imped-
ance depends on the elements’ arrangement. The individual impedances can be connected in
series or in parallel, as shown in Figure 4.2.3.

Series Connection
A closed electrical circuit that contains a power source and two loads connected in series,
with impedances Z1 and Z2 respectively, is presented in Figure 4.2.3(a). According to the law
of conservation of charge, the currents flowing through impedances Z1 and Z2 are the same:
I1 ðsÞ ¼ I2 ðsÞ ¼ I ðsÞ. Also, according to the law of conservation of energy the algebraic sum
of voltages in a closed circuit equals zero, i.e. all the energy generated by the source must be
dissipated or stored by the loads. As illustrated in Section 4.1.1, the voltage gain across
a source and the voltage drop across a load have opposite signs. Assigning a positive
sign to voltage gain and a negative sign to voltage drop, the law of conservation of energy
232 4 Electrical Systems

V 1(s) V 2(s) I(s)

Z 1(s) Z 2(s) e1
+ + V 1(s) V 2(s)

V I 1(s) I 2(s) V I 1(s) Z 1(s) Z 2(s) I 2(s)

– –
I(s) e2

I(s)
(a) (b)

Figure 4.2.3 (a) Series and (b) parallel connections of impedances

for the circuit shown in Figure 4.2.3(a) translates into: V ðsÞ  V1 ðsÞ  V2 ðsÞ ¼ 0 or
V1 ðsÞ þ V2 ðsÞ ¼ V ðsÞ.
Then the equivalent impedance of this circuit is derived as follows:

VðsÞ V1 ðsÞ þ V2 ðsÞ

Zeq ðsÞ ¼ ¼ ¼ Z1 ðsÞ þ Z2 ðsÞ (4.2.6)
IðsÞ IðsÞ

Consequently, a series connection is characterized by the same current passing through the
connected electrical elements. The equivalent impedance for n circuit elements connected in
series is a sum of individual impedances:

X
n
Zeq ðsÞ ¼ Zi ðsÞ (4.2.7)
i¼1

Parallel Connection
A closed electrical circuit that contains a power source and two loads connected in parallel,
with impedances Z1 and Z2 respectively, is presented in Figure 4.2.3(b). According to the law
of conservation of charge the currents flowing through impedances Z1 and Z2 are:
I1 ðsÞ þ I2 ðsÞ ¼ I ðsÞ. The voltage drops across the loads Z1 and Z2 are the same:
V1 ðsÞ ¼ e1  e2 ¼ V2 ðsÞ ¼ V ðsÞ. Then the equivalent impedance is derived as follows:

VðsÞ VðsÞ
Zeq ðsÞ ¼ ¼
IðsÞ I1 ðsÞ þ I2 ðsÞ
(4.2.8)
1 I1 ðsÞ I2 ðsÞ 1 1
¼ þ ¼ þ
Zeq ðsÞ VðsÞ VðsÞ Z1 ðsÞ Z2 ðsÞ

Consequently, parallel connection is characterized by the same voltage drop across the
connected electrical elements. The equivalent impedance for n circuit elements connected in
parallel is computed as:
 4.2 Concept of Impedance 233

1 Xn
1
¼ (4.2.9)
Zeq ðsÞ Z
i¼1 i
ðsÞ

Elements connected either in series or in parallel do not have to be of the same type for the
computation of equivalent impedance. The following examples illustrate the use of Eqs.
(4.2.7) and (4.2.9).

Example 4.2.1 Deriving equivalent impedances for combinations of circuit elements

The connections of several elements and derivations of impedances of these assemblies are shown
in Table 4.2.1, where entries (a) and (b) denote series connections, (c) and (d) denote parallel
connection, and (e) shows a combination of series and parallel connections.

Table 4.2.1 Equivalent impedances

Zeq ¼ Z1 þ Z2 ¼ R1 þ R2

1
Zeq ¼ Z1 þ Z2 þ Z3 ¼ R þ Ls þ
Cs

1 1 1 1 1
¼ þ ¼ þ
Zeq ðsÞ Z1 ðsÞ Z2 ðsÞ R1 ðsÞ R2 ðsÞ

R1 ðsÞR2 ðsÞ
Zeq ðsÞ ¼
R1 ðsÞ þ R2 ðsÞ

1 1 1 1 1
¼ þ ¼ þ
Zeq ðsÞ Z1 ðsÞ Z2 ðsÞ RðsÞ Ls
RLs
Zeq ðsÞ ¼
R þ Ls
Zeq ¼ Zparallel þ ZC
RLs
Zparallel ¼
R þ Ls

RLs 1 RLCs2 þ Ls þ R
Zeq ¼ þ ¼
R þ Ls Cs ðR þ LsÞCs
234 4 Electrical Systems

Example 4.2.2 Finding the current through a circuit element using equivalent
impedances
To illustrate how the use of equivalent impedance simplifies circuit analysis, consider the problem
of finding the current through the inductor L and voltage v1 at terminals A and B for the circuit
shown in Figure 4.2.4.
Solution
Derive an expression for equivalent impedance for the elements connected in parallel:

1 1 1
¼ þ
ZP R 1=Cs
R
ZP ¼
1 þ RCs

According to the law of conservation of energy:

V ðsÞ  VL ðsÞ  VP ðsÞ ¼ 0

where V ðsÞ denotes the voltage supplied by the source, VL ðsÞ is the voltage drop across the
inductor, and VP ðsÞ represents the voltage drop across the impedance ZP .
Using the definition of impedance (Eq. (4.2.1)) and the Laplace transform of the derivative form
of the inductor’s voltage–current relation, we obtain

R
VðsÞ ¼ IðsÞLs þ IðsÞZP ¼ IðsÞ Ls þ
1 þ RCs

The current through the inductor – I ðsÞ – is then derived as

1 þ RCs
IðsÞ ¼ VðsÞ
RLCs2 þ Ls þ R

L
A

I(s)
+
V R C V 1 (s)

–
I(s)

Figure 4.2.4 Circuit for Example 4.2.2

 4.2 Concept of Impedance 235

The voltage at terminals A and B equals the voltage drop across the connected in parallel
component with impedance ZP and is

R
V1 ðsÞ ¼ IðsÞZP ¼ VðsÞ
RLCs2 þ Ls þ R

4.2.3 Divider Rules

The concept of impedance proves very useful for the derivation of two basic rules: the voltage
divider and current divider.
The voltage divider rule is best demonstrated on the series connection of two impedances.
Consider a circuit shown in Figure 4.2.5. The currents through impedances Z1 and Z2
respectively are:

I1 ðsÞ ¼ I2 ðsÞ ¼ IðsÞ (4.2.10)

Using the definition of voltage drop as a difference between electric potentials on the
element’s entrance and exit terminals, we obtain

V1 e A  eB
I1 ¼ ¼
Z1 Z1
(4.2.11)
V2 e B  eC
I2 ¼ ¼
Z2 Z2

The terminals e1 and C are located on the wire that is connected to the ground; hence,
electric potentials of these terminals equal zero. Terminals e2 and A belong to the same wire;
thus, considering the assumption of all wires being ideal, electric potentials of these terminals

I(s)
A

+ e2
I 1 (s) Z 1 (s)

Source
of B
energy
V(s) I 2 (s)
Z 2 (s)
– e1
C
I(s)

Figure 4.2.5 Voltage divider rule

236 4 Electrical Systems

are the same. Also, assigning a positive sign to voltage gain of the energy source as described
in Section 4.2.2, we obtain

VðsÞ ¼ e2  e1 ¼ e2 (4.2.12)

Then currents through impedances Z1 and Z2 respectively are

V  eB eB
I1 ¼ ¼ I2 ¼ ¼ IðsÞ (4.2.13)
Z1 Z2

potential of the terminal B is

Z2
eB ¼ V (4.2.14)
Z1 þ Z2

and the voltage drops across impedances Z1 and Z2 are derived as

Z1
V1 ¼ V
Z1 þ Z2
(4.2.15)
Z2
V2 ¼ V
Z1 þ Z2

The voltage divider rule is expressed as:

V1 Z1
¼ (4.2.16)
V2 Z2

Hence, the voltage drop across the individual impedance – the member of a series
connection – is proportional to the ratio of that impedance and equivalent impedance of
the series connection as a whole. For a circuit that contains a single energy source and n
impedances connected in series, the voltage drop across the i-th impedance is

Zi
Vi ¼ V (4.2.17)
Zeq

where Vi denotes voltage drop across the i-th impedance, V is voltage gain of the source, and
P
n
Zeq ¼ Zi is the equivalent impedance of the circuit.
i¼1
Note that Eq. (4.2.16) can also be derived using the law of conservation of charge and the
definition of impedance. According to the law of conservation of charge, the currents
through impedances Z1 and Z2 are the same:

I1 ðsÞ ¼ I2 ðsÞ ¼ IðsÞ

 4.2 Concept of Impedance 237

By the definition of impedance (Eq. (4.2.1)):

V1 V2
I1 ¼ ¼ I2 ¼
Z1 Z2

And thus
V1 Z1
¼ :
V2 Z2

The current divider rule is best demonstrated on the parallel connection of two impedances.
Consider a circuit shown in Figure 4.2.6. The voltage drops across the impedances Z1 and Z2
are the same:

V1 ¼ V2 ¼ e A  e B (4.2.18)

As described earlier in this section, electric potentials at terminals e1 and B are zero, and
electric potentials at terminals e2 and A are equal:

e2 ¼ eA ¼ VðsÞ:

Thus:

V1 ¼ V2 ¼ VðsÞ (4.2.19)

Using the definition of impedance and Eq. (4.2.19), we obtain

V1 V
I1 ¼ ¼
Z1 Z1
(4.2.20)
V2 V
I2 ¼ ¼
Z2 Z2

Therefore, the current divider rule can be expressed as follows:

I1 Z 2
¼ (4.2.21)
I2 Z 1

I(s)
+ e2 A
V 1 (s) V 2 (s)
Source
of I 1 (s) Z 1 (s) Z 2 (s) I 2 (s)
energy
I(s)
– e1 I(s)
B

Figure 4.2.6 Current divider rule

238 4 Electrical Systems

According to the law of conservation of charge the currents flowing through impedances
Z1 and Z2 are: I1 ðsÞ þ I2 ðsÞ ¼ I ðsÞ. Using this equation together with Eq. (4.2.21), the
expressions for currents across the impedances Z1 and Z2 respectively, are derived as

Z2
I1 ¼ I
Z1 þ Z2
(4.2.22)
Z1
I2 ¼ I
Z1 þ Z2

where I denotes the current through a circuit supplied by the source of current.

Example 4.2.3 Using voltage and current divider rules to find the current through
a specific impedance and the voltage drop across it
To illustrate the use of the voltage and current divider rules in circuit analysis, consider the circuit
shown in Figure 4.2.7(a). We need to find the current through the impedance Z3 and the voltage
drop across it.
Impedances Z2 and Z3 are connected in parallel; thus, their respective voltages are the same as
V2 ¼ V3 ¼ eA . According to the current divider rule, their respective currents are:

Z3
I2 ¼ I
Z2 þ Z3
and
Z2
I3 ¼ I
Z2 þ Z3
where I is the current across the circuit. From the circuit schematic it is obvious that I ¼ I1 . From
V
the definition of impedance, the expression for the current is easily obtained as I ¼ , where V is
Z
the voltage supplied by the energy source, and Z is the equivalent impedance of the circuit.
Let Zeq be the equivalent impedance for the sub-circuit consisting of the impedances Z2 and Z3,
which are connected in parallel. Considering that the circuit is now transformed as shown in
Figure 4.2.7(b), its equivalent impedance is

V 1 (s)
V 1 (s)
A
Z 1 (s) A
Z 1 (s)
I 1 (s) I 3 (s)
+ I 2 (s) V 3 (s) + I(s) I(s) V 3 (s)

V(s) V 2 (s) Z 2 (s) Z 3 (s) V(s) Z eq(s)

– – I(s)
I(s)
(b)
(a)

Figure 4.2.7 Simplifying the circuit using equivalent impedances (see text for details)
 4.2 Concept of Impedance 239

Z ¼ Z1 þ Zeq ;
where
Z2 Z3
Zeq ¼ :
Z2 þ Z3

Hence, the current across the circuit is derived as

Z2 þ Z3
I¼V
Z1 Z2 þ Z1 Z3 þ Z2 Z3

Substituting this result into the expression for the sought current through the impedance Z3 we
find that
Z2
I3 ¼ V :
Z1 Z2 þ Z1 Z3 þ Z2 Z3
Applying the voltage divider rule to the transformed circuit, shown in Figure 4.2.7(b), the
expression for the sought voltage drop across the impedance Z3 is derived as

Zeq Z2 Z3
V3 ¼ Veq ¼ V ¼V :
Z1 þ Zeq Z1 Z2 þ Z1 Z3 þ Z2 Z3

Example 4.2.4 Voltage divider rule in a potentiometer

The operation of a popular electrical device – a potentiometer – is governed by the voltage divider
rule. A potentiometer is a manually adjustable variable resistor with three terminals. Two of the
terminals are connected to the ends of the resistor, and the third one is a sliding electrical pick-off
or wiper. The position of the wiper determines the device’s output voltage V0 . Potentiometers are
used as linear and angular position sensors in various applications. For example, the volume
control knob on the older-model radios is a potentiometer that modifies voltage to the speakers,

V0
R1
wiper I(s)
θmax
wiper
+
+ θ
R1 R0
V(s)
V(s) x R0 V0
L –
– I(s)
I(s)

(a) (b)

Figure 4.2.8 (a) Linear and (b) rotational potentiometers

240 4 Electrical Systems

thus changing the sound volume. Potentiometers may be linear or rotational, as shown in Figure
4.2.8. The resistance R0 between the ground and the wiper is a function of the distance x for linear
potentiometer (see Figure 4.2.8(a)), or of the rotation angle θ for rotational potentiometer (see
Figure 4.2.8(b)).
The voltage output of a potentiometer is computed by applying the voltage divider rule. For the
devices shown in Figure 4.2.8, the circuit’s resistance equals ðR0 þ R1 Þ, and the output voltage is

R0
V0 ¼ V
R 0 þ R1
where V is the voltage supplied by the source of energy, and R0 and R1 are the partial resistances of
the potentiometer as illustrated.
For a linear potentiometer, the resistance R0 is proportional to the device’s length L as follows:

x
R 0 ¼ ð R 0 þ R1 Þ
L
Similarly, for a rotational potentiometer:

θ
R 0 ¼ ð R0 þ R1 Þ :
θmax
Hence, voltage output is derived as
x
V0 ¼ V ¼ KL x for a linear potentiometer
L
θ
V0 ¼ V ¼ Kθ θ for a rotational potentiometer
θmax
The coefficients KL and Kθ are called the potentiometer gain. They depend on the applied voltage
and on the potentiometer geometry, such as the resistor length for a linear potentiometer, and the
wiper’s maximum rotation angle for a rotational potentiometer.
Given the following numerical values for the resistances at a specific position of the wiper:
R0 ¼ 500 Ω and R1 ¼ 1 kΩ, and the input voltage V ¼ 9 V, we find the voltage output as a percent
of the input:

R0 500
V0 ¼ V ¼9 ¼3V
R 0 þ R1 500 þ 1000
which is approximately 33% of the input voltage.

Example 4.2.5
While application of the voltage divider rule simplifies circuit analysis, it needs to be used with
care.
Consider the problem of finding the voltage drop on the resistor R for the circuit shown in
Figure 4.2.9. At the first glance, this circuit carries a strong resemblance to the potentiometer
shown in Figure 4.2.8(a), and one may be tempted to just use the expression for V0 derived in the
previous example. That would be a mistake since the expression for the voltage output of the
 4.2 Concept of Impedance 241

linear potentiometer does not take into consideration the influence of the load R. To account for
that influence, the equivalent resistance Req for the connected in parallel resistors R and R0 needs to
be found prior to applying Eq. (4.2.15), where Z1 ¼ R1 and Z2 ¼ Req .
The equivalent resistance is found as

R0 R
Req ¼
R0 þ R
and the sought voltage drop is then computed as
Req R0 R
V0 ¼ V ¼V :
R1 þ Req R1 R0 þ RR0 þ RR1
This result demonstrates that the addition of a load resistor R to a given potentiometer circuit
with fixed wiper position decreases the output voltage by the amount of

R1 R0 2
V0_no_load  V0_with_load ¼ V :
ðR1 þ R0 ÞðR1 R0 þ RR0 þ RR1 Þ
The current through the circuit with the load increases due to the decrease of the circuit’s
equivalent resistance:
R0 R
R1 þ < R 1 þ R0
R0 þ R

Obviously, the greater the value of R the more pronounced these effects are.
To see the effect of loading the potentiometer wiper use the same values of resistances
R0 ¼ 500 Ω, R1 ¼ 1 kΩ, and the same input voltage V ¼ 9 V as in Example 4.2.4, we compute
the output voltage after adding the load resistor R ¼ 3 kΩ:
R0 R 1000  3000
V0 ¼ V ¼9 ¼ 5:4 V;
R1 R0 þ RR0 þ RR1 500  1000 þ 3000  1000 þ 3000  500

which constitutes 60% of the input instead of 33% for the regular potentiometer without the
loaded wiper.
R

V0

+ R1 R0
V(s)

– I(s)

Figure 4.2.9 Finding voltage drop on the resistor R

242 4 Electrical Systems

4.3 Kirchhoff’s Laws

Kirchhoff’s current law (KCL) and Kirchhoff’s voltage law (KVL) constitute the fundamental
laws of electric circuit modeling. They are derived from the conservation laws of physics.
KCL descends from the law of conservation of charge, while KVL descends from the law of
conservation of energy.
KCL states that for every node of an elec- i1
trical circuit an algebraic sum of currents i2 Node i k+1
equals zero, i.e. the sum of incoming currents
i k+2
equals the sum of outgoing currents.
The mathematical expression of KCL for a ik
in
node that has n current elements, from which
k currents are incoming (see Figure 4.3.1), is Figure 4.3.1 Kirchhoff’s current law

X
n
ij ¼ 0 (4.3.1)
j¼1

where ij indicates an individual current.

Assigning a positive sign to an incoming current, and a negative sign to an outgoing
current, Eq. (4.3.1) may be re-written as

X
k X
n X
k X
n
ij  ij ¼ 0 or ij ¼ ij (4.3.2)
j¼1 j¼k j¼1 j¼k

KVL states that the algebraic sum of volt-

ages around a closed circuit equals zero, i.e. the V 1 (s) V 2 (s)
sum of the voltage gains equals the sum of the Z 1 (s) Z 2 (s)
voltage drops. Alternatively, KVL, being an
e1 I(s) e2
expression of the law of conservation of energy
V n (s)
for electrical circuits, is sometimes formulated I(s)
as: all the energy produced in a closed circuit Z n (s)
must be absorbed by the circuit elements.
The mathematical expression of KVL for a Figure 4.3.2 Kirchhoff’s voltage law
closed circuit that contains n impedances (see Figure 4.3.2), is

X
n
Vcircuit ¼ e1  e2 ¼ Vj ¼ 0 (4.3.3)
j¼1

where Vj indicates the voltage drop across the j-th impedance.

 4.3 Kirchhoff’s Laws 243

Example 4.3.1
For a circuit with a single energy source, the application of Kirchhoff’s laws is mostly straight-
forward. Consider the circuit shown in Figure 4.3.3.
Applying KCL to the node C, for which I1 is the incoming current, and I2 is the outgoing one,
we obtain

I1  I2 ¼ 0 or I1 ¼ I2
Applying KCL to the node B, for which I2 is the incoming current, and I (the circuit current) is
the outgoing one, we obtain
I2  I ¼ 0 or I2 ¼ I
Thus, the same current runs through every element of this circuit:
I1 ¼ I 2 ¼ I
The direction of this current is clockwise according to the definition of current as flow of the
positively charged particles.
In this circuit, a single voltage source generates energy, and two impedances absorb it. In
accordance with convention, a voltage gain associated with energy generation is positive, and a
voltage drop associated with energy consumption is negative. Then, applying KVL to this circuit,
we obtain

V  V1  V2 ¼ 0
Thus, the system of equations that describes this circuit is
I1 ¼ I 2 ¼ I
V  V1  V2 ¼ 0
It can be easily solved for currents and voltage drops by using definition of impedance (Eq. (4.2.1)).

V 1 (s)
A C
Z 1 (s)
+ V 2 (s)
I 1 (s)
V(s) I 2 (s) Z 2 (s)
I(s)
–
B

Figure 4.3.3 Circuit for Example 4.3.1

244 4 Electrical Systems

Example 4.3.2
For a circuit with multiple energy sources, determining the direction of a current and the sign of
the voltage across a source often presents a challenge. In such a circuit, the energy source may
absorb energy instead of generating it, which means that this source experiences a negative voltage
drop rather than the expected voltage gain. To illustrate this concept, consider the circuits with
two energy sources: a voltage source and a current source, shown in Figure 4.3.4.
The direction of the current is defined by the current source, and is, as shown, counterclockwise.
The voltage source in the circuit in Figure 4.3.4(a), according to its shown polarity, experiences
voltage gain: e1 < e2 . So, energy is generated, and v is positive.
Also, since all the connecting wires are considered ideal, e4 ¼ e2 > 0 and e3 ¼ e1 < 0. Thus, the
current source experiences a voltage drop: e4 > e3 , which means that this element absorbs energy
and its voltage vi is negative.
For the circuit in Figure 4.3.4(b) the situation is reversed. The voltage source consumes energy,
experiencing a voltage drop ðe1 > e2 Þ, while the current source generates energy, experiencing a
voltage gain ðe4 < e3 Þ. Therefore, v is negative, and vi is positive.
Note that Kirchhoff’s laws can be expressed either in time or in Laplace domains. The choice of
domain depends on the circuit model requirements. Nonetheless, since circuit analysis aims at a
finding circuit response, the voltage–current relation expression in the time domain is customarily
sought.

– e1 e3 + e1 e3

i i
v v
vi vi
+ e2 e4 – e2 e4

(a) (b)

Figure 4.3.4 Circuits with two energy sources (see text for details)

4.4 Passive-Circuit Analysis

The straightforward application of the divider rules and Kirchhoff’s laws to a single-loop
electrical circuits yields the voltage–current relations for each circuit element with relative
ease, as demonstrated earlier in this chapter. For multiple-loop circuits, such a direct
approach may fail to derive the correct model. Hence, several methods were developed to
simplify the complex multi-loop circuits. While the detailed discussion on the circuit analysis
methodology belongs to the field of electrical engineering and is beyond the scope of this
book, an introduction to the most widely used approaches is presented.
 4.4 Passive-Circuit Analysis 245

The methods explored further in this chapter include: (a) the loop method, (b) the node
method, (c) the superposition theorem, (d) Thevenin’s theorem, and (e) Norton’s theorem.

4.4.1 Loop Method

The loop method of circuit analysis utilizes the concept of loop currents, which are the main
parameters of the circuit model. A loop current is a current assigned for a specific loop in the
circuit. The loop method encompasses the following steps.
First, currents for every loop in the circuit are assigned, using either the marked polarity of
the voltage power source, or the current direction indicated by the current power source. For
interconnecting elements – elements that simultaneously belong to more than a single loop –
net currents are used. A net current is an algebraic sum of all the currents passing through the
interconnecting element. In evaluation of the element’s net current, the direction of the
assigned loop current in the examined loop is defined as positive.
After the currents passing through every circuit element are defined, the expressions for
the voltages are derived using Ohm’s law, the known voltage–current relations, and the
divider rules.
The last step of this method includes the application of KVL to every loop in the circuit.
The derived KVL equations comprise the sought system model in terms of loop currents as
independent variables. The output equation is also formulated in terms of loop currents.

Example 4.4.1 To illustrate operation of the loop method, consider the circuit shown in Figure
4.4.1. This circuit consists of four passive elements: two resistors R1 and R2 , two capacitors C1 and
C2 , and a single source of energy. The polarity of the voltage source is as shown on the circuit
schematic. The system input is the voltage vin supplied by the voltage source, and the output is the
capacitor C2 voltage vout . Zero initial conditions are assumed.
As indicated in the circuit schematic, this circuit consists of two loops, for which the loop
currents i1 and i2 are defined as shown. The direction of i1 is clockwise due to the marked polarity
of the voltage source. Consequently, the direction of i2 is also clockwise.
This circuit has a single interconnecting element: capacitor C1 . Its net current is i1  i2 for loop
1, and i2  i1 for loop 2.
Applying KVL to loop 1 yields

vin  vR1  vC1 ¼ 0

and to Loop 2:
v C1 þ v R 2 þ v C2 ¼ 0
Recalling the voltage–current relations for resistor (Eq. (4.1.9)) and capacitor (Eq. (4.1.16)), the
KVL equations become
246 4 Electrical Systems

R1 R2

i1 i2
+

vin C1 C2 vout
i1 i2
– i2

Loop 1 Loop 2

Figure 4.4.1 Modeling a passive circuit using the loop method

ðt
1
Loop 1 vin ðtÞ ¼ R1 i1 þ ði1  i2 Þdt (1)
C1
0
ðt ðt
1 1
Loop 2 R2 i2 þ ði2  i1 Þdt þ i2 dt ¼ 0 (2)
C1 C2
0 0

The output equation is derived as

ðt
1
vout ðtÞ ¼ i2 dt (3)
C2
0

These three equations constitute the required circuit model.

If desired, this model can be transformed into a single differential equation that directly
expresses the relation between the input and output voltages. The easiest way to achieve such a
transformation is by taking the Laplace transform of Eqs. (1), (2), and (3), performing the
necessary algebraic operations, and taking the inverse Laplace transform of the resulting equa-
tion. This derivation is as follows (capital letters denote variables in Laplace domain):

1
Vin ¼ R1 I1 þ ðI1  I2 Þ (4)
C1 s
1 1
R 2 I2 þ ðI2  I1 Þ þ I2 ¼ 0 (5)
C1 s C2 s
1
Vout ¼ I2 (6)
C2 s

From Eq. (6):

I2 ¼ C2 sVout (7)
 4.4 Passive-Circuit Analysis 247

Then, substituting this result into Eq. (5), we obtain the expression for I1 :

1 1
I1 ¼ C1 s R 2 þ þ C2 sVout (8)
C1 s C2 s
Finally, substituting Eqs. (7) and (8) into Eq. (4), we derive the sought model:
 
1 1 1 1
Vin ¼ R1 þ R2 þ
þ C1 C2 s2 Vout  C2 sVout
C1 s C1 s C2 s C1 s (9)
 
¼ R1 R2 C1 C2 s2 þ ðR1 C1 þ R2 C2 þ R1 C2 Þs þ 1 Vout

Taking the inverse Laplace transform of Eq. (9), the system model is obtained as a differential
equation in terms of input and output voltages:

d2 vout dvout
vin ¼ R1 R2 C1 C2 þ ð R 1 C1 þ R 2 C2 þ R 1 C2 Þ þ vout (10)
dt2 dt

4.4.2 Node Method

The node method of circuit analysis operates with the concept of node voltages, which are the
main parameters of the circuit model. A node is a point in an electric circuit at which
connecting wires intersect or branch. The node voltage is an electric potential assigned to a
node, and an independent variable in the equations that constitute the circuit model. The
node method operates on the circuit as a whole, without subdividing it into loops, and
encompasses the following steps.
First, the nodes of interest and their electric potentials – node voltages – are defined. Then,
the current for each circuit element is assigned, using the indicated polarity of the voltage
source or marked current direction of the current source.
The expressions for the current through each circuit element are written in terms of node
voltages, using the known voltage–current relations, Ohm’s law, and the divider rules.
After all the expressions for component currents are derived, KCL is applied to each
node. These KCL equations comprise the sought system model in terms of node
voltages as independent variables. The output equation is also formulated in terms of
node voltages.

Example 4.4.2
To illustrate the operation of the node method, consider the circuit used in the Example 4.4.1.
As shown in Figure 4.4.2, three nodes are defined, with assigned node voltages e0 , e1 , and e2
respectively, and the element currents are as indicatedon the circuit schematic.
248 4 Electrical Systems

Remembering that the electric potential of any point on a bottom wire is zero (this is the wire
connected to the ground) and using the known voltage–current relations for the resistor and
capacitor the element currents in terms of node voltages are derived as:
e0  e1 vin  e1
iR1 ¼ ¼
R1 R1
e1  e2 e1  vout
iR2 ¼ ¼
R2 R2
d de1
iC 1 ¼ C1 ðe1  0Þ ¼ C1
dt dt
d de2 dvout
i C 2 ¼ C2 ðe2  0Þ ¼ C2 ¼ C2
dt dt dt

R1 R2
#0 #1 #2
e0 iC e1 iC e2
+ 1 2
iR iR vout
vin 1 2 C2
C1
–

Figure 4.4.2 Modeling a passive circuit using the node method

Applying KCL to nodes 1 and 2 we obtain

Node 1 iR1  iC1  iR2 ¼ 0
Node 2 iR2  iC2 ¼ 0
Substituting the derived element current expressions into KCL equations, we obtain the
required system model:

vin  e1 de1 e1  vout

 C1  ¼0
R1 dt R2
e1  vout dvout
 C2 ¼0
R2 dt
It is a system of two differential equations with two variables: e1 and vout . Since vout is an output,
no separate output equation is needed.
If desired, the intermediate variable e1 can be eliminated, thus reducing the system to a single
differential equation of a single independent variable vout . Expressing e1 in terms of vout , using the
second equation of the above system model yields
dvout
e 1 ¼ R 2 C2 þ vout
dt
Substituting this result into the first equation of the derived system model, we obtain
 4.4 Passive-Circuit Analysis 249

  
vin 1 1 dvout d2 vout dvout vout
 þ R 2 C2 þ vout  C 1 R 2 C2 þ þ ¼0
R1 R1 R2 dt dt2 dt R2
Algebraic transformations result in the following system model:
d2 vout dvout
R 1 R 2 C1 C2 2
þ ðR1 C1 þ R2 C2 þ R1 C2 Þ þ vout ¼ vin
dt dt
As expected, this model is identical to that derived in the Example 4.4.1 (see Eq. (10)).

4.4.3 Circuits with Multiple Energy Sources

Straightforward applications of the loop method and the node method to the analysis of a
circuit with multiple energy sources may prove difficult. The Superposition theorem is one of
the approaches that allow efficient dealing with the multiple energy sources within a circuit.
Thevenin’s theorem and Norton’s theorem constitute the other popular approaches to the
simplification of such a circuit. The usability of these three methods is particularly apparent
for the equivalent transformation of a circuit that is the combination of resistors and sources.

Superposition Theorem
Superposition theorem states that the current through any passive element of a circuit with
multiple energy sources equals the algebraic sum of the currents produced by each individual
source independently. To evaluate the current resulting from the i-th source, all the other
sources are “turned off.” This means replacing the voltage sources by short circuits (zero
voltage), and the current sources, by open circuits (zero current). Then, either of the
described earlier methods for circuit analysis can be applied.

Example 4.4.3
Consider the circuit shown in Figure 4.4.3(a). It contains two energy sources – a current source
I ðsÞ and a voltage source V ðsÞ. This problem can be solved either in the time domain or in the
Laplace domain. Here, the Laplace domain solution is provided, which is easily inverted into the
time domain when required.
We want to evaluate the total current passing through the resistor R1 . According to the
superposition theorem, this current (I1 ) is a sum of two current components – one due to the
current source (II1 ), and the other due to the voltage source (IV1 ):

I1 ¼ IV1 þ II1
To derive the current IV1 resulting from the voltage source, the current source is turned off as
shown in Figure 4.4.3(b). Since the right loop of the circuit depicted in Figure 4.4.3(b) is open,
there is no current there. Thus, only the left loop consisting of the voltage source and two resistors
R1 and R3 participates in current generation. The resistors R1 and R3 are connected in series, so
250 4 Electrical Systems

currents flowing through these elements are the same: IV1 ¼ IV3 ¼ IV . The current IV1 is then
derived using Ohm’s law and the equivalent resistance of the circuit:

V V
IV1 ¼ ¼
RVeq R1 þ R3
To derive the current II1 resulting from the current source, the voltage source is turned off as
shown in Figure 4.4.3(c). Here both loops of the circuit have a current.
According to the law of conservation of charge, the current flowing through the resistor R2
equals the current supplied by the energy source. The resistors R1 and R3 are connected in parallel
so, using the current divider law (Eq. (4.2.22)), the current across the resistor R1 is
R3
II1 ¼ I
R1 þ R 3
Note that the currents II1 and IV1 have opposite directions. Assuming the clockwise direction to
be positive, the total current across the resistor R1 is

V R3
I1 ¼ IV1 þ II1 ¼ I
R1 þ R3 R1 þ R 3

+ R1 R2 I(s)

V(s) R3

(a)
I v 1(s) I I1(s) I I (s)
2

+ R1 R2 R1 R2 I(s)
V(s)
R3 I I3(s) R3
– I v (s) I I(s)

(b) (c)

Figure 4.4.3 Modeling a passive circuit using the superposition theorem (see text for details)

Thevenin’s and Norton’s Theorems

Thevenin’s theorem states that any combination of energy sources and resistors with two
terminals can be replaced by an equivalent circuit of a single voltage source VT (Thevenin
 4.4 Passive-Circuit Analysis 251

voltage) and a single series resistor RT (Thevenin resistance). The Thevenin voltage is the
open-circuit voltage at the terminals. The Thevenin resistance is the resistance of the open
circuit measured at the terminals. Its value is obtained by evaluating the equivalent resistance
of this circuit, where voltage sources are replaced by short circuits, and current sources are
replaced by open circuits.
Norton’s theorem states that any combination of energy sources and resistors with two
terminals can be replaced by an equivalent circuit of a single current source IN (Norton
current) and a single resistor RN (Norton resistance) in parallel with the current source. The
Norton resistance is defined in the same way as the Thevenin resistance. The Norton current
is the current that would have existed in the wire connecting the terminals of the open circuit.
Its value is obtained by Ohm’s law, dividing the open-circuit voltage at the terminals by the
Norton resistance. If the chosen terminals for the Norton and Thevenin transformations are
the same, then RN ¼ RT and IN ¼ VT =RT .
An illustration of the equivalent transformation of an electrical circuit using Thevenin’s
and Norton’s theorems is presented in the example below.

Example 4.4.4 Consider the two-loop circuit shown in Figure 4.4.4(a). Thevenin’s and Norton’s
equivalents of the right-hand side of this circuit at the terminals A and B are constructed as shown
in Figure 4.4.4(b) and Figure 4.4.4(c) respectively.
As stated earlier, RT ¼ RN . To find its value consider the open circuit shown surrounded with a
dashed line in Figure 4.4.4(a). Replacing the voltage source V1 with a short circuit, it can be easily

A A A

R0 R1 R3 R0 RT R0

+ –
V1 RN IN
+ + – +
V0 V0 VT V0
R2
– – + –

B B B
Thevenin’s Norton’s
equivalence equivalence
(a) (b) (c)

Figure 4.4.4 Modeling a passive circuit using Thevenin’s and Norton’s theorems (see text for details)
252 4 Electrical Systems

seen that resistors R1 and R3 are connected in parallel, and resistor R2 is in series with their
equivalence. Thus:

R1 R3
RT ¼ R N ¼ R 2 þ
R 1 þ R3
The Thevenin voltage is the open-circuit voltage at the terminals A and B. This voltage equals
the voltage drop across the resistor R1 . The resistor R2 has no influence on the Thevenin voltage
since there is no current passing through it. To find this voltage consider the loop, consisting of the
voltage source V1 and resistors R1 and R3 . In this loop, the resistors are connected in series, so the
voltage divider rule is used (Eq. (4.2.15)):

R1
VT ¼ V 1
R 1 þ R3
Then, the Norton current is
VT R1
IN ¼ ¼ V1
RT R 1 R 2 þ R1 R 3 þ R 2 R 3
The polarity of the Thevenin’s voltage source and the direction of the current provided by the
Norton’s current source are defined to maintain the direction of current in the transformed
equivalent circuit as the same as it was before the transformation.

Example 4.4.5
This example encompasses the application of the circuit analysis methods described in this section
to a simple two-loop electrical circuit. The consistency of the derived results is demonstrated.
Consider the two-loop circuit shown in Figure 4.4.5.
We need to derive a circuit model and find the current and the voltage drop for the resistor R3 , if
the numerical values of circuit elements are as follows:

R1 ¼ 5 Ω; R2 ¼ 3 Ω; R3 ¼ 2 Ω;
v1 ¼ 12 V; v2 ¼ 8 V

+ R1 R3 +
v1 R2 v2

– –

Figure 4.4.5 Circuit for Example 4.4.5

 4.4 Passive-Circuit Analysis 253

Solution by the Loop Method

The loops are defined and the loop currents are assigned in accordance with the indicated polarity
of the voltage sources as shown in Figure 4.4.6.
Resistor R2 is the interconnecting element and its net current is the same for both loops:
iR2 ¼ i1 þ i2 . Recalling the voltage–current relation of a resistor (Eq. (4.1.9)), apply KVL to
loop 1 and loop 2:

Loop 1 v1  i1 R1  ði1 þ i2 ÞR2 ¼ 0

Loop 2 v2  i2 R3  ði1 þ i2 ÞR2 ¼ 0
This system of two equations with two independent variables – i1 and i2 – constitutes the system
model.
Since the only current that passes through R3 is i2 , we find the required current and voltage by
solving the above system of equations for i2 and then, using Ohm’s law, we derive vR3 :

R2 v1 þ ðR1 þ R2 Þv2

iR 3 ¼ i 2 ¼ ¼ 0:9032 A
R 1 R 2 þ R 1 R 3 þ R2 R 3
vR3 ¼ iR3 R3 ¼ 1:8064 V
Considering that system inputs are voltages v1 and v2 provided by the voltage sources, and the
outputs are the current and the voltage drop for the resistor R3 , the system model can be written in
terms of input–output as follows:
R2 v1 þ ðR1 þ R2 Þv2
System model iR 3 ¼
R 1 R 2 þ R 1 R 3 þ R2 R 3
Output equations fiR3 ; vR3 ¼ iR3 R3 g

Solution Using Thevenin’s Theorem

This solution uses an equivalent transformation of the circuit. We define the terminals A and B as
shown in Figure 4.4.7(a), and substitute the left-hand side of the circuit shown encircled in a
dashed rectangle with its Thevenin equivalent.

R1 R3

R2
+ i1 +
v1 i1 i2 v2
– i2 –
Loop 1 Loop 2

Figure 4.4.6 Solution by the loop method

254 4 Electrical Systems

A A

+ R1 R3 + + RT R3 +
v1 R2 v2 VT v2
i2
– – – –
B B
Thevenin’s
equivalence
(a) (b)

Figure 4.4.7 Solution by Thevenin’s theorem (see text for details)

The Thevenin resistance is computed as equivalent resistance for two resistors R1 and R2
connected in parallel:
R1 R2
RT ¼
R 1 þ R2
The Thevenin voltage equals the voltage drop across the resistor R2 and is derived using the
voltage divider rule (Eq. (4.2.15)):

R2
VT ¼ v1 :
R 1 þ R2
The equivalent single-loop circuit is shown in Figure 4.4.7(b). Since the resistors RT and R3 are
connected in series, the current passing through them is the same and can be derived by direct
application of Ohm’s law. Considering the direction of this current to be as shown on the circuit
schematic, we can state that the voltage source v2 generates energy, i.e. experiences a voltage gain,
while the Thevenin voltage source VT is absorbing energy, i.e. experiences a voltage drop. Thus, v2
is positive, and VT is negative. Then, the system model is

v2  VT ¼ iðRT þ R3 Þ
or in extended form:

R2 R1 R2
v2  v 1 ¼i þ R3
R1 þ R 2 R 1 þ R2
Then, the required current and voltage for the resistor R3 are found as:
R2
v2  v1
R1 þ R 2 ðR1 þ R2 Þv2  R2 v1
iR 3 ¼ i ¼ ¼ ¼ 0:9032 A
R1 R2 R 1 R 2 þ R1 R 3 þ R2 R 3
þ R3
R 1 þ R2
vR3 ¼ iR3 R3 ¼ 1:8064 V
Considering the specified inputs and outputs of the system, the system model can be defined in
terms of input–output as follows:
 4.4 Passive-Circuit Analysis 255

ðR1 þ R2 Þv2  R2 v1
System model iR 3 ¼
R 1 R 2 þ R 1 R 3 þ R2 R 3
Output equations fiR3 ; vR3 ¼ iR3 R3 g

Solution Using Norton’s Theorem

This solution also uses an equivalent transformation of the circuit. We define the terminals A and
B as shown in Figure 4.4.8(a), and substitute the part of the circuit shown encircled with a dashed
line with its Norton equivalent. The equivalent circuit is shown in Figure 4.4.8(b).
Even though the terminals chosen for the Norton transformation are not the same as those
chosen for the Thevenin transformation in the previous solution, the Norton resistance is the same
as the Thevenin resistance:

R1 R2
RN ¼
R 1 þ R2
To derive the Norton current, the voltage at the terminals A and B needs to be evaluated first.
From the circuit schematic in Figure 4.4.8(a) it is clearly seen that this voltage is a difference
between the voltage gain of the source v2 and the voltage drop across the resistor R2 , which is
found using the voltage divider rule:

R2
vAB ¼ v2  vR2 ¼ v2  v1
R1 þ R 2
Hence, Norton current is derived as
R2
v2  v1
vAB R1 þ R2 ðR1 þ R2 Þv2  R2 v1
iN ¼ ¼ ¼
RN R1 R2 R1 R2
R 1 þ R2
The resulting Norton-equivalence circuit has a single current source and two resistors con-
nected in parallel. Then, the current through the resistor R3 is found using current divider rule:

A B
A B R3
i2
+ R1 R3 +
v1 R2 v2
RN
IN
– –

Norton’s equivalence
(a) (b)

Figure 4.4.8 Solution by Norton’s theorem (see text for details)

256 4 Electrical Systems

R1 R2
RN ðR1 þ R2 Þv2  R2 v1 R1 þ R2 ðR1 þ R2 Þv2  R2 v1
iR3 ¼ iN ¼ ¼ ¼ 0:9032 A
R N þ R3 R1 R2 R R
1 2 R 1 R 2 þ R1 R 3 þ R 2 R 3
þ R3
R 1 þ R2

and the voltage drop is computed using Ohm’s law:

vR3 ¼ iR3 R33 ¼ 1:8064 V
Considering the specified inputs and outputs of the system, the system model can be defined in
terms of input–output as follows:
ðR1 þ R2 Þv2  R2 v1
System model iR 3 ¼
R 1 R 2 þ R1 R 3 þ R2 R 3
Output equations fiR3 ; vR3 ¼ iR3 R3 g

Solution Using the Superposition Theorem

Since the Superposition theorem deals with currents through circuit elements, the system model
will be derived in the form
iR3 ¼ i13 þ i23
where iR3 is the current through the resistor R3 , i13 is the current through this resistor due to the
voltage source v1 , and i23 is the current through this resistor due to the voltage source v2 . Note
that as discussed earlier in this section, the sum of the component currents is algebraic,
which means that the signs of the component currents i13 and i23 are assigned according to
their directions.
The directions of currents i13 and i23 are defined as shown in Figure 4.4.9.
To find the current i13 , the voltage source v2 is replaced by a short circuit as shown in Figure
4.4.9(a).
Then, the equivalent resistance of this circuit is

R1 R3 R1 R3

+ i 23 +
i 13
v1 i1 R2 R2 v2
i2
– –

(a) (b)

Figure 4.4.9 Solution by the Superposition theorem

 4.4 Passive-Circuit Analysis 257

R2 R3
Req ¼ R1 þ
R 2 þ R3

and the circuit current is:

v1 v1 ðR2 þ R3 Þ
iv 1 ¼ ¼
R2 R3 R 1 R 2 þ R1 R 3 þ R2 R 3
R1 þ
R 2 þ R3
R2 R3
Since resistors R1 and the equivalence are in series, the same current passes through them.
R 2 þ R3
Then, i13 is found using the current divider rule:

R2 v 1 R2
i13 ¼ iv1 ¼
R 2 þ R 3 R 1 R 2 þ R1 R 3 þ R2 R 3
Similarly, i23 is found as the circuit current for the circuit shown in Figure 4.4.9(b):
v2 v 2 ð R 1 þ R2 Þ
i23 ¼ iv2 ¼ ¼
R1 R2 R1 R 2 þ R1 R 3 þ R2 R 3
R3 þ
R1 þ R 2
The currents i13 and i23 have opposite directions, so they must have different signs in the
algebraic sum that represents the current through the resistor R3 . Assume that i23 is positive.
Then, the required current and voltage for the resistor R3 are found as:

ðR1 þ R2 Þv2  R2 v1
iR3 ¼ i23  i13 ¼ ¼ 0:9032 A
R 1 R 2 þ R1 R 3 þ R2 R 3

vR3 ¼ iR3 R3 ¼ 1:8064 V

The computed result for iR3 is positive, which means that the positive current component was
chosen correctly. A negative result would have indicated that the original assumption of the
direction of current through R3 is incorrect, and the current actually flows in the opposite
direction.
The derived expressions for current and voltage for the resistor R3 constitute the system model
in terms of input and output:
ðR1 þ R2 Þv2  R2 v1
system model i R3 ¼
R 1 R 2 þ R 1 R 3 þ R2 R 3
output equations fi R 3 ; v R 3 ¼ i R 3 R 3 g
As expected, the results derived by all the methods employed in this example are identical.
258 4 Electrical Systems

4.5 State-Space Representations and Block Diagrams

4.5.1 Block Diagrams
As discussed in Section 2.7.2, block diagrams and transfer functions are intimately related. A
block diagram is an extremely good tool for describing the input–output relations for a
variety of dynamic systems, and electrical circuits in particular. While a circuit schematic
depicts the physical connections between the system components, a block diagram deals with
variables being transformed and propagated through the system. As such, it is a useful visual
medium aimed at better understanding of the relations between the elements of a dynamic
system.
The generic block diagram is shown in Figure 4.5.1.
Customarily, all the elements of a block diagram are in the Input Transfer Output
function
Laplace domain.
The presented block diagram may refer to the dynamic Figure 4.5.1 Generic
system as a whole, in which case it can be viewed as a diagram- block diagram
matic extension of the system transfer function model. It may
also refer to an individual system element. Then the transfer-
function block describes the element’s dynamics, while the input and output are the inter-
mediate variables of the system.
Consider the simple single-loop electrical circuit shown in
+ R
Figure 4.5.2. Let the input be the voltage vin , supplied by the C
vout
voltage source, and the output be the voltage drop vout on the vin
i
capacitor C. –
The construction of a block diagram for this circuit encom-
passes the following steps: (a) deriving the system model equa-
tions in time domain, (b) taking the Laplace transform of the
Figure 4.5.2 Simple
system equations considering zero initial conditions, (c) deriv- RC circuit
ing the component transfer functions and constructing their
graphical representation – component block diagrams, as
required to comprehensively express the interactions between circuit elements, and (d)
assembling the system block diagram from the constructed components.
The equation that constitutes the circuit model is derived by applying KVL:
ð
1
vin  vR  vC ¼ vin  iR  idt ¼ 0 (4.5.1)
C

The output equation expresses the voltage–current relation for a capacitor:

ð
1
vout ¼ idt (4.5.2)
C

Taking the Laplace transform of Eqs. (4.5.1) and (4.5.2), we obtain

 4.5 State-Space Representations and Block Diagrams 259

1 RCs þ 1
Vin ðsÞ ¼ IðsÞR þ IðsÞ ¼ IðsÞ (4.5.3)
Cs Cs
1
Vout ðsÞ ¼ IðsÞ (4.5.4)
Cs

The sought system block diagram is in the form shown in Figure 4.5.1, where Vin ðsÞ is the
system input, and Vout ðsÞ is the system output. The circuit current I ðsÞ is the intermediate
variable. Equation (4.5.3) describes the relation between the variables Vin ðsÞ and I ðsÞ, where
the former is the component input, the latter is the component output, and the transfer
Cs
function is . Similarly, for Eq. (4.5.4) I ðsÞ is the component input, Vout ðsÞ is the
RCs þ 1
1
component output, and the transfer function is .
Cs
The component block diagrams are shown in Figure 4.5.3.
The system block diagram is constructed by combining these component block diagrams;
see Figure 4.5.4(a). The intermediate variable I ðsÞ can be removed by the algebraic trans-
formation of this block diagram, shown in Figure 4.5.4(b).
As seen from the final system diagram (Figure 4.5.4(b)), the system transfer function is
1
.
RCs þ 1
For circuits with multiple energy sources, the construction of a block diagram is more
involved. The complexity of a block diagram depends on the circuit inputs and outputs. The
procedure is similar to that described in the Example 4.5.1: (a) the equations relating inputs
to outputs are derived using the relevant circuit analysis method, (b) the Laplace transform
of them is taken, (c) block diagram components are built from individual equations, and
finally (d) the components are combined into a finished block diagram.

V in (s) Cs I(s) I(s) 1 Vout (s)

RCs+1 Cs

(a) (b)

Figure 4.5.3 Component block diagrams for (a) voltage and (b) current

V in (s) Cs I(s) 1 Vout(s) V in (s) 1 Vout(s)

RCs+1 Cs RCs+1

(a) (b)

Figure 4.5.4 Construction of a system block diagram (see text for details)
260 4 Electrical Systems

Example 4.5.1
Consider the circuit shown in Figure 4.5.5. It consists of a current source and three passive
elements: a resistor R, a capacitor C, and an inductor L. Construct a block diagram of this circuit,
considering that the input is represented by the current is , and the outputs are voltages vR , vC , and
vL across the circuit components.
This problem is approached using the node method. The nodes 1 and 2 are defined as shown in
the circuit schematic. Applying KCL to them, we obtain

Node 1 iS ¼ i R þ i L (1)
Node 2 iL ¼ i C (2)
The next step involves expressing the currents in Eqs. (1) and (2) in terms of appropriate voltages.
From Eq. (1):

iR ¼ i S  i L (3)
Then the voltage across the resistor is
vR ¼ RiR ¼ RðiS  iL Þ (4)
The voltage across the inductor is
diL
v L ¼ v R  vC ¼ L (5)
dt
From Eq. (5) the current through the inductor is
ð
1
iL ¼ ðvR  vC Þdt (6)
L
Substituting this result into Eq. (4), we obtain
 ð
1
vR ¼ R iS  ðvR  vC Þdt (7)
L
The current through the capacitor is
dvC
iC ¼ C (8)
dt

L
#1 #2

is R iL

iR iC

Figure 4.5.5 RLC circuit for Example 4.5.1

 4.5 State-Space Representations and Block Diagrams 261

Substituting the results of Eqs. (6) and (8) into Eq. (2), and taking the derivative with respect to
time in order to get rid of the integral, we obtain
ð
dvC 1
iC ¼ C ¼ iL ¼ ðvR  vC Þdt
dt L (9)
d 2 vC
LC 2 ¼ vR  vC
dt
Hence, the equations needed for construction of a block diagram are (5), (7), and (9). They fully
express the relations between the specified input and outputs.
Taking the Laplace transform of these equations, we obtain
VL ðsÞ ¼ VR ðsÞ  VC ðsÞ (10)

1 
VR ðsÞ ¼ R IS ðsÞ  VR ðsÞ  VC ðsÞ (11)
Ls
VR ðsÞ  VC ðsÞ ¼ LCs2 VC ðsÞ (12)
Rearranging the terms in these equations to make the construction of the block diagram easier:
VL ðsÞ ¼ VR ðsÞ  VC ðsÞ (13)

RLs 1
VR ðsÞ ¼ IS ðsÞ þ VC ðsÞ (14)
Ls þ R Ls
1
VC ðsÞ ¼ VR ðsÞ (15)
LCs2 þ 1
The resulting block diagram is shown in Figure 4.5.6(a). It is one of the possible block diagrams,
but hardly the most straightforward one due to the relatively complicated transfer functions of its
elements.
A simpler and more intuitive block diagram can be constructed using Eqs. (2), (4), (5), and (8) in
the following way:
ð
1
vL ¼ vR  vC ¼ iL dt
L
vR ¼ RðiS  iL Þ (16)
dvC
iC ¼ C ¼ iL
dt
Taking the Laplace transform of Eq. (16), we obtain
VL ðsÞ ¼ VR ðsÞ  VC ðsÞ ¼ Ls IL ðsÞ
VR ðsÞ ¼ RðIS ðsÞ  IL ðsÞÞ (17)
IC ðsÞ ¼ CsVC ðsÞ ¼ IL ðsÞ
The resulting block diagram is shown in Figure 4.5.6(b).
Deriving the transfer function for every output is straightforward. For example, the transfer
function for the capacitor is derived from the block diagram in Figure 4.5.6(b) as follows:
262 4 Electrical Systems

VR (s)
+
– VL (s)
Is (s) + RLs 1
– Ls+R VR (s) LCs 2+1 VC (s)

1 VC (s) Inputs
Ls

(a)
VR (s)

VL (s)

Is (s) + IR (s) + IL (s) = IC (s) 1 VC (s)

1
R
– VR (s) – VL (s) Ls Cs
IL (s) Outputs

(b)

Figure 4.5.6 Block diagrams for the RLC circuit of Example 4.5.1 (see text for details)

  1 1
IS ðsÞ  IL ðsÞ R  VC ðsÞ ¼ VC ðsÞ
Ls Cs
   
IS ðsÞ  IL ðsÞ R ¼ LCs2 þ 1 VC ðsÞ

Since IL ðsÞ ¼ CsVC ðsÞ, we obtain

 
RIS ðsÞ ¼ LCs2 þ RCs þ 1 VC ðsÞ
VC ðsÞ R
¼ (18)
IS ðsÞ LCs2 þ RCs þ 1
To demonstrate the correctness of this transfer function, the differential equation for the
voltage across the capacitor is derived using Eqs. (7) and (9):
 ð
d 2 vC 1 d 2 vC
LC 2 þ vC ¼ R iS  LC 2 dt
dt L dt (19)
2
d vC dvC
LC 2 þ RC þ vC ¼ RiS
dt dt
It is clearly seen that Eq. (19) is consistent with the transfer function depicted by Eq. (18).
 4.5 State-Space Representations and Block Diagrams 263

Example 4.5.2
Consider the circuit shown in Figure 4.5.7. It consists of two voltage source and three passive
elements: resistor R, capacitor C, and inductor L.
A block diagram of this circuit can be constructed, considering that the input is represented by
the voltages v1 and v2 , and the outputs are the currents through the circuit components iR , iC , and
iL .
The loop method is used to derive the circuit model. The loop currents i1 and i2 are assigned as
shown in Figure 4.5.7. The currents through the circuit components are expressed in terms of loop
currents as follows:
iR ¼ i 1 (1)
iL ¼ i 1 þ i 2 ¼ i R þ i C (2)
iC ¼ i 2 (3)

Applying KVL to the circuit loops, we obtain

diL
v1 ¼ v R þ v L ¼ i R R þ L (4)
dt
ð
1 diL
v 2 ¼ v C þ vL ¼ iC dt þ L (5)
C dt
Taking the Laplace transform of Eqs. (2), (4), and (5) yields the relations necessary for
constructing individual components of the sought block diagram:
V1 ðsÞ ¼ IR ðsÞR þ LsIL ðsÞ (6)
1
V2 ðsÞ ¼ IC ðsÞ þ LsIL ðsÞ (7)
Cs
IL ðsÞ ¼ IR ðsÞ þ IC ðsÞ (8)
Rearranging Eqs. (6) and (7) to have currents as outputs, the block diagram components are
constructed as shown in Figure 4.5.8:

1 
IR ðsÞ ¼ V1 ðsÞ  LsIL ðsÞ (9)
R

R C

+ i1 +
v1 i1 v2
i2
– –
i2

Loop 1 Loop 2
Figure 4.5.7 RLC circuit with multiple energy sources
264 4 Electrical Systems

V1 (s) + 1 IR (s) V2 (s) + IC (s)

Cs IR (s)
– R – + IL (s)

IL (s) IL (s) +
Ls Ls IC (s)

(a) (b) (c)

Figure 4.5.8 Component block diagrams for the circuit of Example 4.5.2: (a) corresponds to Eq. (9),
(b) to Eq. (10), and (c) corresponds to Eq. (8)

V1 (s) + IR (s) IR (s)

1
– R

IL (s) + IL (s)
Ls
+

V2 (s) – IC (s) IC (s)

Cs
+
Outputs

Figure 4.5.9 Block diagram for the circuit of Example 4.5.2

 
IC ðsÞ ¼ Cs V2 ðsÞ  LsIL ðsÞ (10)

Combining these components yields the system block diagram as shown in Figure 4.5.9.
For verification purposes, the transfer functions for the current IL ðsÞ are derived using the block
diagram above, and compared with the differential equation where iL ðtÞ is an independent
variable.
From the block diagram, the expression for IL ðsÞ is derived as follows:
1   
IL ðsÞ ¼ V1 ðsÞ  LsIL ðsÞ þ Cs V2 ðsÞ  LsIL ðsÞ
R (11)
 
LRCs þ Ls þ R IL ðsÞ ¼ V1 ðsÞ þ RCsV2 ðsÞ
2

From Eq. (11), the transfer functions are easily obtainable as follows:

IL ðsÞ 1
¼ (12)
V1 ðsÞ LRCs þ Ls þ R
2

IL ðsÞ RCs
¼ (13)
V2 ðsÞ LRCs2 þ Ls þ R
 4.5 State-Space Representations and Block Diagrams 265

To derive the sought differential equation in terms of iL ðtÞ, we derive the expression for iR ðtÞ
from Eq. (4), and the expression for iC ðtÞ from Eq. (2):

1 diL
iR ¼ v1  L (14)
R dt

1 diL
iC ¼ iL  iR ¼ iL  v1  L (15)
R dt
We substitute the result of Eq. (15) into Eq. (5) and take a derivative with respect to time to get
rid of the integral:

dv2 1 d 2 iL 1 1 diL d 2 iL
¼ iC þ L 2 ¼ i L  v1  L þL 2 (16)
dt C dt C RC dt dt
Then, the system model in terms of the current through the inductor as an independent variable is
d 2 iL diL dv2
LRC þL þ RiL ¼ v1 þ RC (17)
dt2 dt dt
It is obvious that the Eq. (17) is consistent with the transfer functions described with Eqs. (12)
and (13).

4.5.2 State-Space Representation

Similarly to the other dynamic systems, an electric circuit can be modeled with
differential equations, transfer functions, block diagrams, and state-space representa-
tion. Generally, an electric circuit has multiple current and voltage variables as well as
more than a single energy source. An incorrectly selected set of independent variables
may lead to an unduly complicated model, and, subsequently, difficulties in simulation
and analysis of the circuit.
A state-space representation, first introduced in Section 2.5.4, is a way of modeling an
electrical circuit simply and efficiently. A linear state-space model can be derived for the
circuits discussed in this chapter. These state equations are a set of coupled first-order
ordinary differential equations of the form represented by Eq. (2.5.10).
The state variables, capable of comprehensively modeling an electrical circuit, describe the
energy stored by the circuit elements. Every energy-storing element is assigned a single state
variable.
As described in Section 4.1.2, the only elements that store energy are inductors and
capacitors. The energy-storing capability is evidenced by the presence of a derivative in
their voltage–current relations. Hence, the number of state variables for a circuit equals the
number of inductors and capacitors in that circuit. Resistors do not get a state variable since
they don’t store any energy.
266 4 Electrical Systems

The form of a linear state equation, where the left-hand side contains the first derivative of
a state variable, and the right-hand side does not have any derivatives of state variables and
inputs, prompts the choice of the state variables. Examining the derivative form of the
voltage–current relation for a capacitor (Eq. (4.1.16)), the assignment of a state variable to
obtain the linear state equation is easy – it is a parameter that has the first derivative taken,
which is the voltage. Thus, the state variable x and the corresponding state equation for a
capacitor are:

x ¼ vðtÞ
1 (4.5.5)
x_ ¼ i ðtÞ
C

Similarly, for an inductor, the current is selected a state variable (Eq. (4.1.21)):

x ¼ i ðtÞ
1 (4.5.6)
x_ ¼ vðtÞ
L

Upon the assignment of state variables, any of the presented earlier circuit analysis
methods can be used to derive the governing equations. It is customary to present the circuit
state-space model, described by the state and output equations, in matrix form.
The following examples illustrate the derivation of a state-space model for electrical
circuits.

Example 4.5.3
Derive the state-space model of a circuit shown in Figure 4.5.10, and present it in matrix form. The
voltage vin , supplied by the energy source, is the system input, and voltage drop vout across the
capacitor is the output.

Solution
This circuit needs two state variables since it has two energy-storing elements – one inductor L and
one capacitor C. Therefore, as discussed above, the state variables are: the current i through an
inductor, and the voltage v across the capacitor:
x1 ¼ i (1)
x2 ¼ vout (2)
Before applying KVL to this circuit, we write the voltage–current relations for its components
in terms of the above state variables:

Resistor vR ¼ iR ¼ x1 R (3)
 4.5 State-Space Representations and Block Diagrams 267

R L

+
vout
vin i C

Figure 4.5.10 RLC circuit for Example 4.5.3

ð ð
1 1
Capacitor vC ¼ vout ¼ x2 ¼ idt ¼ x1 dt (4)
C C
dx2 1
¼ x_ 2 ¼ x1 (5)
dt C
di
Inductor vL ¼ L ¼ Lx_ 1 (6)
dt
Applying KVL to the circuit, we obtain
di
vin ¼ vR þ vL þ vC ¼ iR þ L þ vout (7)
dt
In terms of state variables Eq. (7) becomes
vin ¼ x1 R þ Lx_ 1 þ x2 (8)
Then, the state-space model of this circuit is represented as follows:
8
> R 1 1
>
<x_ 1 ¼  L x1  L x2 þ L vin
State equations
>
> 1
: x_ 2 ¼ x1
C
Output equation y ¼ x2 (9)
In matrix form

X_ ¼ A  X þ B  u
(10)
y¼CX
where the components are:
x1
State vector X¼
x2
Input ðscalarÞ u ¼ vin
2 R 13 2 3
  1
6 L7
Matrices A¼4 L 5; B ¼ 4L 5; C ¼ ½0 1 
1
0 0
C
268 4 Electrical Systems

Example 4.5.4
Derive the state-space model of a dual-resistance circuit, discussed in Example 4.4.2, and present
it in matrix form. The voltage vin , supplied by the energy source, is the system input, and the
voltage drop vout across the capacitor C2 is the output.
This circuit has two energy-storing components – capacitors C1 and C2 . Hence, it needs two
state variables, defined as:
x1 ¼ vC1 ¼ e1
(1)
x2 ¼ vC2 ¼ vout
The previously derived component equations are expressed in terms of the state variables
defined by Eq. (1):
vin  e1 vin  x1
iR1 ¼ ¼ (2)
R1 R1
e1  vout x1  x2
iR2 ¼ ¼ (3)
R2 R2
de1 dx1
iC1 ¼ C1 ¼ C1 ¼ C1 x_ 1 (4)
dt dt
dvout dx2
iC2 ¼ C2 ¼ C2 ¼ C2 x_ 2 (5)
dt dt
Re-writing the derived KCL equations in terms of state variables, we obtain
vin  x1 dx1 x1  x2
 C1  ¼0 (6)
R1 dt R2
x1  x2 dx2
 C2 ¼0 (7)
R2 dt
Then, the state-space model of this circuit is derived as follows:
8 
>
> 1 vin  x1 x1  x2
<x_ 1 ¼ C
>
R1

R2
1
State equations
>
> x1  x2
>
: x_ 2 ¼
R2 C 2
Output equation y ¼ x2 (8)
In matrix form:

X_ ¼ A  X þ B  u
(9)
y¼CX
 4.5 State-Space Representations and Block Diagrams 269

where the components are:

x1
State vector X¼
x2
Input ðscalarÞ u ¼ vin
2 3
R 1 þ R2 1 2 3
6 R1 R2 C1 R 2 C1 7
1
6 7
Matrices A¼6 7; B ¼ 4R1 C1 5;
4 1 5
 1 0
R 2 C2 R 2 C2
C ¼ ½0 1 

Example 4.5.5
Derive the state-space model of a circuit shown in Figure 4.5.11, and present it in matrix form. The
voltage vin , supplied by the energy source, is the system input, and the voltage drop vout across the
capacitor C2 is the output.
This circuit needs three state variables since it has three energy-storing elements – one inductor
L and two capacitors C1 and C2 . Therefore:
x 1 ¼ iL
x2 ¼ vC1 ¼ e2 (1)
x3 ¼ vC2 ¼ e3 ¼ vout
Note that e1 – voltage at the node 1 – is an intermediate variable, which will be eliminated
during the derivation of governing equations.
The most efficient way to derive a state-space representation of this circuit is by using the node
method. The nodes are defined as shown in the circuit schematic in Figure 4.5.11. The expressions
for the current through every circuit element are derived in terms of state variables as follows:

R1 L R2
#1 #2 #3
e1 e2 e3
+
iR iL iR
1 2
C2 vout
vin C1

– iC iC
1 2

Figure 4.5.11 Circuit for Example 4.5.5

270 4 Electrical Systems

vin  e1
iR1 ¼ iL ¼ x1 ¼ (2)
R1
e2  e3 e2  vout x2  x3
iR2 ¼ ¼ ¼ (3)
R2 R2 R2
de2 dx2
iC1 ¼ C1 ¼ C1 ¼ C1 x_ 2 (4)
dt dt
dvout dx3
iC2 ¼ C2 ¼ C2 ¼ C2 x_ 3 (5)
dt dt
From Eq. (2):
e1 ¼ vin  R1 x1 (6)
Also, from Eq. (6) and the voltage–current relation for an inductor:
diL 1 1
¼ ðe1  e2 Þ ¼ x_ 1 ¼ ðvin  R1 x1  x2 Þ (7)
dt L L
Apply KCL to the identified nodes:
Node 1 iR1 ¼ iL (8)
Node 2 iL ¼ iC1 þ iR2 (9)
Node 2 iR2 ¼ iC2 (10)
Combining the element equations with the node equations and performing the necessary
algebraic transformations, the state-space model is derived:
8
>
> R1 1 1
> x_ 1 ¼  x1  x2 þ vin
>
>
> L L L
>
< 1 1 1
State equations x_ 2 ¼ x1  x2 þ x3
>
> C1 R 2 C1 R 2 C1
>
>
>
> 1 1
>
: x_ 3 ¼ x2  x3
R 2 C2 R 2 C2
Output equation y ¼ x3 (11)
In matrix form:
X_ ¼ A  X þ B  u
(12)
y¼CX
where the components are:
" #
x1
State vector X ¼ x2
x3
Input ðscalarÞ u ¼ vin
2 R 1 3
1
  0 2 3
6 L L 7 1
6 7
6 1 1 7 6 L7
A¼6  1 7
Matrices 6 R 2 C1 R2 C1 7; B ¼ 40 5;
6 C1 7
4 1 1 5 0
0 
R 2 C2 R 2 C2
C ¼ ½0 0 1 
 4.6 Passive Filters 271

4.6 Passive Filters

Filters are important component in the modern electrical and electronic devices. They form
an indispensable component of any signal-processing hardware/software assembly, perform-
ing a variety of operations such as conditioning signals prior to analog-to-digital conversion,
smoothing data sets, blurring of images, reducing high- and low-frequency noise in acoustic
applications, and many others. Filters can be represented by passive and active circuits. This
section deals with the passive-circuit implementation of common filters, such as low-pass,
high-pass, and band-pass filters.

4.6.1 Low-Pass Filter

A low-pass filter is a passive circuit, the purpose of which is to pass signals with a frequency
lower than the designed cutoff frequency, and attenuate signals with higher than cutoff
frequencies. Two types of this filter are shown in Figure 4.6.1.
To derive the transfer function for these filters and analyze how they attenuate high-
frequency signals, use the impedance method.
The transfer function for an RL filter is derived using KVL:

Vin ðsÞ ¼ VL þ VR ¼ LsI þ RI ¼ IðLs þ RÞ (4.6.1)

Since the output vout is the voltage drop across the resistor, we find

Vout
I¼
R

Substituting this expression into Eq. (4.6.1), we derive the transfer function

Vout ðsÞ R
GRL ðsÞ ¼ ¼ (4.6.2)
Vin ðsÞ Ls þ R

L R

+ +
vin R vout vin C vout
i i
– –

(a) (b)

Figure 4.6.1 Passive low-pass filters: (a) series RL filter, (b) series RC filter
272 4 Electrical Systems

Converting this transfer function to a frequency domain (refer to Chapter 7 for a detailed
discussion on frequency response) and finding its magnitude, we obtain




Vout ð jωÞ
1
jGRL ð jωÞj ¼


¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

(4.6.3)
Vin ð jωÞ 1 þ L2 R2 ω2

The cutoff frequency is defined as the frequency at which VR ¼ VL , and is found as

1
ω0 ¼
RL

Using the same reasoning, derive the transfer function for the RC filter:

1 RCs þ 1
Vin ðsÞ ¼ VC þ VR ¼ I þ RI ¼ I
Cs Cs
1 1
Vout ðsÞ ¼ I ) GRC ðsÞ ¼ (4.6.4)
Cs

RCs þ 1

Vout ð jωÞ
1


jGRC ð jωÞj ¼

¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Vin ð jωÞ
1 þ C2 R2 ω2

and its cutoff frequency is

1
ω0 ¼ :
RC

To illustrate attenuation of signals with frequency greater than ω0 generate Bode plots
using the BodePlot Mathematica function (see Figure 4.6.2).
For both filters, the resistance was the same: R ¼ 1 kΩ, and the values of the inductance
and capacitance varied as shown in Figure 4.6.2.
As derived in Chapter 7, the amplitude of the frequency response of a dynamic system
subject to a periodic input with magnitude A that equals AjGð jωÞj (see Eqs. (7.3.5) and
(7.3.6)). The gain (jGð jωÞj) equals 1 for all input frequencies less than or equal to the cutoff
frequency ω0 , which means that these inputs will be passed through the filter without a
decrease in amplitude. For signal frequencies greater than ω0 the amplitude of the response
will be decreasing rapidly, approaching zero – this signifies signal attenuation.
The cutoff frequency for both filters increases with the decrease of inductance or capaci-
tance: the lowest ω0 were registered for the highest inductance L ¼ 200 mH and for the
highest capacitance C ¼ 200 nF.
Mathematica code that generates the Bode plot for RL filter is shown in Figure 4.6.3.

4.6.2 High-Pass Filter

A high-pass filter is a passive circuit, the purpose of which is to pass signals with a frequency
higher than the designed cutoff frequency, and attenuate signals with lower than cutoff
frequencies. Two types of this filter are shown in Figure 4.6.4.
 4.6 Passive Filters 273

Figure 4.6.2 Bode plots (gain only) for low-pass passive filters: (a) RL filter, (b) RC filter

Using the same reasoning as in Section 4.6.1, we can derive the transfer functions and
generate Bode plots for these filters:




Vout ð jωÞ
LRω


jGRL ð jωÞj ¼

¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (4.6.5)
Vin ð jωÞ
1 þ L2 R2 ω2
274 4 Electrical Systems

Figure 4.6.3 Mathematica code for generating the Bode gain plot for the RL filter

R
C

+ +
vin vout vin vout
i R i L
– –

(a) (b)

Figure 4.6.4 Passive high-pass filters: (a) series RC filter, (b) series RL filter

Vout ð jωÞ
Cω
jGRC ð jωÞj ¼


¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (4.6.6)
Vin ð jωÞ
1 þ C2 R2 ω2

The Bode plots for these high-pass filters are shown in Figure 4.6.5.
The attenuation of low-frequency signals is clearly seen in these graphs. The same
tendency is found: the cutoff frequency ω0 increases with decreasing inductance or capaci-
tance; and the lowest ω0 was registered for the highest inductance L ¼ 200 mH and for the
highest capacitance C ¼ 200 nF.

4.6.3 Band-Pass Filter

A band-pass filter is a passive circuit, the purpose of which is to pass signals of the frequency
being within the specific limits ω01 ≤ ω ≤ ω02 , where ω01 and ω02 are the lower and upper
cutoff frequencies. It is represented by an RLC circuit as shown in Figure 4.6.6.
 4.6 Passive Filters 275

Figure 4.6.5 Bode plots (gain only) for high-pass passive filters: (a) RL filter, (b) RC filter
276 4 Electrical Systems

L C

+
vin vout
i R
–

Figure 4.6.6 Band-pass passive filter

The KVL and impedance method is used to derive the transfer function of this circuit:

1 LCs2 þ RCs þ 1
Vin ðsÞ ¼ VC þ VL þ VR ¼ I þ LsI þ RI ¼ I
Cs Cs
RCs
Vout ðsÞ ¼ RI ) GRLC ðsÞ ¼
LCs2 þ RCs þ 1




Vout ð jωÞ
ωðR=LÞ
jGRLC ð jωÞj ¼


¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (4.6.7)
Vin ð jωÞ

1 2 
R 2
 ω2 þ ω
LC L

The maximum value of the transfer function is jGRL ð jωÞj ¼ 1. The cutoff frequencies are
1
defined as those at which jGRL ð jωÞj ¼ pffiffiffi. Solving this equation, we obtain the values of
2
cutoff frequencies:
s ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
R R 2 1
ω01 ¼ þ þ
2L 2L LC
s
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (4.6.8)
R R 2 1
ω02 ¼ þ þ
2L 2L LC

The bandwidth is the difference between the cutoff frequencies:

R
ω01  ω02 ¼ (4.6.9)
L

Using the following numeric values for the system parameters: R ¼ 1 kΩ; C ¼ 2 nF;
L ¼ 2 μH, the Bode plot shown in Figure 4.6.7 is generated.
This band-pass filter will pass the signals with frequencies between 0.16 MHz and 0.16 GHz,
and will attenuate all the others.
 4.7 Active-Circuit Analysis 277

Figure 4.6.7 Bode plot of a band-pass passive filter

4.7 Active-Circuit Analysis

As discussed earlier in this chapter, a generic passive circuit consists of an energy source and
passive elements such as resistors, capacitors, and inductors. An active circuit also has
passive components that constitute a load, and an amplifier – an integrated circuit that
contains transistors, resistors, and capacitors, and is capable of amplifying (increasing) an
electronic signal without affecting it in any other way.
There exists an extensive classification of amplifiers according to their circuit configur-
ations, type and size of input signal, and frequency of operations. This book is concerned
only with the DC small-signal amplifiers, in particular isolation amplifiers and operational
amplifiers (op-amps).

4.7.1 Isolation Amplifiers

A voltage-isolation amplifier is an integrated circuit, the purpose of which is to amplify a
signal from a low-power source by yielding an output voltage proportional to the input
voltage by a predefined factor. Obviously, it requires an external energy source to provide an
input voltage. Its design must comply with two important requirements: (a) its operation
should not affect the behavior of the energy source circuit, and (b) it should be capable of
providing the required voltage independently of the load circuit.
A voltage-isolation amplifier circuit is shown in Figure 4.7.1.
278 4 Electrical Systems

Zsrc Za_out
A1 Va B1
+ Is Ia
Vsrc Za_in Ga Zload Vload

– A2 B2

Source circuit Amplifier Load circuit

Figure 4.7.1 Voltage isolation amplifier

Here, Vsrc denotes the voltage provided by the energy source, Zsrc and Zload are the
impedances of the source circuit and the load, respectively, Vload is the voltage delivered to
the load, Ga denotes the voltage gain of the amplifier, and Za_in and Za_out are the impedances
on the input and output terminals of the amplifier, respectively. All these parameters are in
the Laplace domain.
The amplifier input terminals denoted by A1 and A2 , and the amplifier input voltage, Va , is
the voltage measured at these terminals.
We let Is be the current through the source circuit impedance Zsrc . Applying KVL to the
source circuit, we obtain

Vsrc  Is Zsrc  Va ¼ 0 (4.7.1)

Considering that the impedances Zsrc and Za_in are connected in series, Is is the current
drawn by the amplifier circuit. It is obtained using Ohm’s law:

Va
Is ¼ (4.7.2)
Za_in

Equation (4.7.2) illustrates that, if the amplifier input impedance is very large, Is will be
negligibly small, meaning that the amplifier does not affect the current through the source
circuit. Hence, to avoid affecting the behavior of the source circuit, an amplifier must have a
large input impedance.
Also, using the voltage divider rule (Eq. (4.2.15)), find that

Za_in
Va ¼ Vsrc ≈ Vsrc (4.7.3)
Zsrc þ Za_in

when the amplifier input impedance is sufficiently large, Za_in >> Zsrc .
The amplifier output voltage, measured at the output terminals B1 and B2 , is found as
 4.7 Active-Circuit Analysis 279

Va_out ¼ Ga Va ≈ Ga Vsrc (4.7.4)

Then, considering that impedances Za_out and Zload are connected in series, the application
of KVL to the load circuit yields

Va_out  Ia Za_out  Ia Zload ¼ 0 (4.7.5)

which derives the following expression for the amplifier current Ia :

Va_out Ga Va Ga Vsrc
Ia ¼ ¼ ≈ (4.7.6)
Za_out þ Zload Za_out þ Zload Za_out þ Zload

Thus, the voltage delivered to the load is

Zload
Vload ¼ Ga Vsrc (4.7.7)
Za_out þ Zload

Equation (4.7.7) shows that if the amplifier output impedance, Za_out , is very small, the
voltage delivered to the load becomes independent of the load circuit: Vload ≈ Ga Vsrc .
Hence, a voltage-isolation amplifier must have a large input impedance and a small output
impedance in order to avoid affecting the source circuit, while providing the required voltage
independently of the load circuit. This voltage is proportional to the voltage, supplied by the
energy source, by a factor of Ga – the amplifier gain. Therefore, a voltage-isolation amplifier
behaves as a voltage source.
Current-isolation amplifiers provide a current, proportional to the input generated by the
energy source. The same essential characteristics apply: the required current is provided
regardless of the load circuit, and the amplifier operation does not affect the energy source.
Thus, a current-isolation amplifier behaves as a current source. A similar approach can be
used to demonstrate that for a current-isolation amplifier:

Iload ≈ Ga Isrc (4.7.8)

Example 4.7.1
Here, we revisit the circuit in the Example 4.4.1, and modify the schematic as shown in Figure
4.7.2, where the amplifier is shown as a simple box with the gain Ga .
In the modified circuit the source and load loops are connected through a voltage-isolation
amplifier instead of directly to each other. The amplifier is assumed to comply with the impedance
requirements discussed above. Consequently, it does not affect the behavior of the source circuit,
while providing the voltage independently of the load circuit. The amplifier then prevents the
source circuit (loop 1 in Example 4.4.1) from influencing the load circuit (loop 2 in the Example
4.4.1), thus, creating two separate independent loops.
Consider the source circuit. The voltage across the capacitor C1 equals the input voltage into the
amplifier, Va_in , and is derived using the voltage divider rule (Eq. (4.2.15)):
280 4 Electrical Systems

R1 R2

A1 B1
+
Isrc Iload
Amplifier C2 Vout
Vsrc
C1 Ga
–
A2 B2

Source circuit Load circuit

Figure 4.7.2 Circuit for Example 4.7.1

1=C1 s 1
VC1 ¼ Vsrc ¼ Vsrc ¼ Va_in (1)
R1 þ 1=C1 s R1 C1 s þ 1

1
Note, that the transfer function of this loop is .
R 1 C1 s þ 1
According to Eq. (4.7.4) the amplifier output voltage is
Va_out ¼ Ga Va_in (2)
Consider the load circuit. The output voltage equals the voltage across the capacitor C2 that is
derived using the voltage divider rule (Eq. (4.2.15)):

1=C2 s 1
Vout ¼ VC2 ¼ Va_out ¼ Va_out (3)
R2 þ 1=C2 s R2 C2 s þ 1
1
Note, that the transfer function of the load loop is .
R 2 C2 s þ 1
Substituting the expressions for the input and output voltages of the amplifier (Eqs. (1) and (2))
into Eq. (3), we obtain the circuit model in Laplace domain:
1 1 1
Vout ¼ Va_out ¼ Ga Vsrc
 R2 C 2 s þ 1 R2 C2 s þ 1 R1 C1 s þ 1 (4)
R1 R2 C1 C2 s2 þ ðR1 C1 þ R2 C2 Þs þ 1 Vout ¼ Ga Vsrc

By taking the inverse Laplace transform of Eq. (4) obtain the system model in the time domain:
d2 vout dvout
Ga vsrc ¼ R1 R2 C1 C2 þ ðR1 C1 þ R2 C2 Þ þ vout (5)
dt2 dt
Comparing the derived governing equation (Eq. (5)) with that obtained for the unmodified
circuit (Eq. (10) of Example 4.4.1), one can clearly see the difference. Even if assuming Ga ¼ 1,
these models are still not identical. As seen from Eq. (4), the transfer function of the circuit with
the isolation amplifier is a product of the transfer functions of two independent circuits – the
source loop and the load loop, while for the circuit without the amplifier it is not (see Eq. (9) of
 4.7 Active-Circuit Analysis 281

Example 4.4.1). This illustrates the influence two dependent loops exert on each other’s voltages
and currents, as presented in Example 4.4.1.
Therefore, in any multiple-loop circuit, where the loops are connected end to end, there exists
the so-called loading effect – the voltage and current in any given loop are affected by the adjacent
loops. An isolation amplifier, placed between loops, eliminates this effect, making the loops
independent of each other.

4.7.2 Operational Amplifiers

An operational amplifier is capable of almost ideal amplification of DC signals, which makes
it one of the major building blocks of analog electronic circuits. In addition to signal
amplification, op-amps are used for signal filtering, and for performing mathematical oper-
ations such as, for example, addition/ subtraction and integration/ differentiation.
While the concept of an op-amp is not new – the original vacuum-tube-based device was
invented at the Bell Laboratories in the early 1940s and became commercially available
approximately 10 years later – the op-amps are still widely used. The most common op-amp
application is as a component of control systems.
An ideal op-amp is a three-terminal device as
shown in Figure 4.7.3. The physical integrated Inputs
Output
circuit (chip) has a relatively complicated struc- v–
– vout
ture and more than three terminals, but this sim-
plified representation is sufficient for analysis +
and, consequently, for representation on elec- v+
trical circuit schematics. ia
A typical linear op-amp, as shown in Figure
4.7.3, is a voltage amplifier with a very large gain. Figure 4.7.3 Ideal inverting op-amp
It has two high-impedance inputs – an inverting
input denoted by v , and a non-inverting input
denoted by vþ . Then, the actual voltage input of this op-amp is as follows:

vin ¼ vþ  v (4.7.9)

Since only linear op-amps are considered, the output voltage is proportional to the input
voltage as follows:

vout ¼ Kvin ¼ K ðvþ  v Þ (4.7.10)

where K is the op-amp gain. For an ideal device, this gain is assumed to be infinite, while for a
physical device it is in the range of 105–106.
282 4 Electrical Systems

The output terminal, vout , can be viewed as an amplification factor, and is often referred to
as the op-amp gain. The output impedance of an ideal op-amp is assumed to be zero, while the
physical devices usually have a small internal resistance in the order of 100–20 Ω.
Due to the very large input impedance, an op-amp draws negligible current:

ia ≈ 0 (4.7.11)

Therefore, the voltage difference between the input terminals approaches zero:
vout
ia ¼ ¼ v þ  v ≈ 0
K (4.7.12)
vþ ≈ v

The expression

vþ ≈ v  (4.7.13)

is customarily called the op-amp equation.

The concept of impedance is instrumental for modeling and analysis of a typical op-amp
circuit, as shown in Figure 4.7.4. Since impedance is traditionally expressed in the Laplace
domain, in the context of this analysis all the voltages, currents, and electric potentials are
also expressed in Laplace domain.
A typical inverting op-amp circuit is presented in Figure 4.7.4, where Z1 indicates the input
impedance, while Z2 denotes the output or feedback impedance; Vin is the input voltage, Vout
is the output voltage, Ein signifies the electric potential on the inverting terminal, and Eout is
the electric potential on the output terminal.
Recalling the voltage–current relation of an impedance (Eq. (4.2.1)):

VðsÞ
IðsÞ ¼ (4.7.14)
Z

Figure 4.7.4 Typical inverting op-amp circuit

 4.7 Active-Circuit Analysis 283

Applying KCL to the node A, we obtain

I1 ¼ I2 þ I3 (4.7.15)

Since an op-amp draws negligible current, I3 ¼ Ia ≈ 0. Then, the currents through the input
and output impedances are equal:

I1 ¼ I2 (4.7.16)

From Eq. (4.6.14), we obtain the expressions for these currents:

Ein  Ea 
I1 ¼
Z1
 (4.7.17)
Ea  Eout
I2 ¼
Z2

and substitute them into Eq. (4.6.16):

Ein  Ea  Ea   Eout
¼ (4.7.18)
Z1 Z2
Considering that the non-inverting input terminal of this op-amp is grounded, the input
and output voltages are, respectively,

Vin ¼ Ein  Ea þ ¼ Ein  0 ¼ Ein

(4.7.19)
Vout ¼ Eout  Ea þ ¼ Eout  0 ¼ Eout

From the op-amp equation (Eq. (4.7.13)):

Ea þ ≈ Ea  ≈ 0 (4.7.20)

Substituting Eqs. (4.7.19) and (4.7.20) into Eq. (4.7.18), we obtain

Vin Vout Z2
¼ or Vout ¼  Vin (4.7.21)
Z1 Z2 Z1

This equation illustrates the concept of the inverting op-amp circuit operation – the amplifi-
cation and inversion of the input voltage.
Depending on the type of basic elements used as impedances Z1 and Z2 , an op-amp circuit
can perform different functions.

4.7.3 Operational Amplifier Circuits

Multiplier (Amplifier) Circuits
The impedances Z1 and Z2 in an op-amp circuit are represented by two resistances Ri and Rf
as shown in Figure 4.7.5.
284 4 Electrical Systems

Figure 4.7.5 Multiplier circuit

The multiplier equation can be obtained by using the same reasoning as for the previous
derivation. Note that for this derivation all the currents and voltages are expressed in the
time domain.
Applying KCL to the node A, we obtain

i 1 ¼ i2 þ i 3 (4.7.22)

Since an op-amp draws negligible current, i3 ¼ ia ≈ 0. Therefore:

i1 ¼ i2 (4.7.23)

Using Eq. (4.1.8) to obtain the expressions for these currents, and substituting them into
Eq. (4.7.23):

ein  ea  ea   eout
¼ (4.7.24)
Ri Rf

Since the non-inverting input terminal of this op-amp is grounded, the input and output
voltages are, respectively,

vin ¼ ein  ea þ ¼ ein  0 ¼ ein

(4.7.25)
vout ¼ eout  ea þ ¼ eout  0 ¼ eout

From the op-amp equation (Eq. (4.7.13)):

e a þ ≈ ea  ≈ 0 (4.7.26)

Substituting Eqs. (4.7.25) and (4.7.26) into Eq. (4.7.24), we obtain

vin vout Rf
¼ or vout ¼  vin (4.7.27)
Ri Rf Ri
 4.7 Active-Circuit Analysis 285

Hence, in a multiplier, the output voltage is proportional to the input voltage by the factor
Rf
, and the sign of the output voltage is inverted.
Ri
The multiplier equation (Eq. (4.7.27)) can be alternatively derived using Eqs. (4.7.21) and
(4.2.3). Since Z1 ¼ Ri and Z2 ¼ Rf , substituting into Eq. (4.7.21) yields

Rf
Vout ¼  Vin
Ri

or in time domain
Rf
vout ¼  vin
Ri

Inverter (Sign-Changer) Circuits

An inverter op-amp circuit is a variation of the
multiplier circuit, where the impedances Z1 and Z2
are represented by two equal resistances R as shown
in Figure 4.7.6. This schematic utilizes the simplified
representation of an inverting op-amp – a triangle
with the minus sign inside. Figure 4.7.6 Inverter circuit
Using Eq. (4.7.27), the inverter equation is easily derived:

R
vout ¼  vin ¼ vin (4.7.28)
R
An inverter circuit does not change the magnitude of an input voltage, but changes its sign.
It is often used as an inverting buffer in logic circuits.

Integrator Circuits
In an integrator op-amp circuit, the impedance Z1 is represented by a resistor R, and Z2 by a
capacitor C, as shown in Figure 4.7.7.
The integrator equation can be obtained by using the same reasoning as before. Recalling
that i1 ¼ i2 , and using the expressions for the voltage–current relation for resistors and
capacitors (Eqs. (4.1.8) and (4.1.16)), we obtain

ein  ea  d
¼ C ðea   eout Þ (4.7.29)
R dt

Furthermore, recalling that vin ¼ ein , vout ¼ eout , and ea þ ≈ ea  ≈ 0:

286 4 Electrical Systems

Figure 4.7.7 Integrator circuit

ðt
vin dvout 1
¼ C or vout ¼ vin dt (4.7.30)
R dt RC
0

Alternatively, the integrator equation (Eq. (4.7.30)) is derived using Eqs. (4.7.21), (4.2.3),
1
and (4.2.4). Since Z1 ¼ R and Z2 ¼ , substituting into Eq. (4.7.21) yields:
Cs
1
Vout ¼  Vin
RCs

or in time domain
ðt
1
vout ¼ vin dt
RC
0

As illustrated by the integrator equation (Eq. (4.7.30)), this circuit provides the output
1
voltage proportional to the time integral of its input voltage by the factor of , and with the
RC
opposite sign. Unlike the multiplier and inverter op-amp circuits, the integrator circuit
causes the output voltage to respond to the changes in the input voltage over time rather
than to the input magnitude.
Due to the presence of a feedback capacitor, a nonideal integrator experiences the
undesirable effect of an output-voltage offset. A capacitor functions as an open circuit
until accumulating sufficient charge or saturating (see Section 4.1.2). Hence, while the
capacitor is charging, the integrator loses its feedback and acts as a regular op-amp,
described by Eq. (4.7.10). That means that even a very small voltage input results in a high
output due to the high gain of an op-amp. This may push the circuit beyond its operating
limits, which is, obviously, unacceptable.
 4.7 Active-Circuit Analysis 287

To combat this weakness, the integrator circuit is

modified as shown in Figure 4.7.8. In electrical engin-
eering, this design is sometimes called a lag or first-
order low-pass filter.
The addition of a resistor Rf parallel to the capacitor
in the feedback line ensures that no feedback is lost
while the capacitor is charging, thus minimizing an
Figure 4.7.8 Lag circuit
output-voltage offset.
The easiest way to derive a governing equation for this op-amp circuit is by using Eq.
(4.7.21) and the known expressions for resistor and capacitor impedances as well as that for
the parallel connection of impedances (Eqs. (4.2.3), (4.2.4), and (4.2.9)):

Z1 ¼ Ri
1 1 1 Rf
¼ þ ) Z2 ¼ (4.7.31)
Z2 Rf 1=Cf s 1 þ Rf C f s

Then, the relation between the input and output voltages in the Laplace domain is

Z2 Rf
Vout ¼  Vin ¼    Vin (4.7.32)
Z1 R i 1 þ Rf Cf s
Rf
Thus, while the capacitor is charging, the output voltage becomes Vout ¼  Vin instead
Ri
of the undesirable large-value Vout ¼ KVin (see Eq. (4.7.10)) as it was for a simple integrator.
Transforming Eq. (4.7.32) to the time domain, we obtain

Rf
Vout þ Rf Cf sVout ¼  Vin
Ri
(4.7.33)
dvout Rf
vout þ Rf Cf ¼  vin
dt Ri

Typically, the relation between resistances of the added feedback component Rf and the
original resistor Ri is Rf > 10Ri .
Another modification of an integrator circuit is called a bandwidth-limited integrator,
shown in Figure 4.7.9.
The governing equation for this circuit is derived using Eq. (4.7.21) and the known
expressions for the resistor and capacitor impedances as well as the one for the series
connection of impedances (Eqs. (4.2.3), (4.2.4), and (4.2.7)):

Z 1 ¼ Ri
1 1 þ Rf Cf s (4.7.34)
Z 2 ¼ Rf þ ¼
Cf s Cf s

Then, the relation between input and output voltages in the Laplace domain is
288 4 Electrical Systems

Figure 4.7.9 Bandwidth-limited integrator circuit

Z2 1 þ Rf Cf s
Vout ¼  Vin ¼  Vin (4.7.35)
Z1 R i Cf s

Transforming Eq. (4.6.35) to the time domain, we obtain


dvout dvin
R i Cf ¼  vin þ Rf Cf (4.7.36)
dt dt

or, integrating once with respect to time and considering zero initial conditions:
ðt
Rf 1
vout ¼  vin  vin dt (4.7.37)
Ri R i Cf
0

Integrator circuits were one of the staples of analog computers. At the present time, they
are mostly used in analog-to-digital converters, and various applications that involve wave
shaping (for example, a square-wave input yields a triangular-wave output, while a sine input
results in a cosine output).

Differentiator Circuits
In a differentiator op-amp circuit, the impedance
Z1 is represented by a capacitor C, and Z2 by a
resistor R, as shown in Figure 4.6.10.
The derivation of its governing equation is
straightforward, using the same reasoning as for Figure 4.7.10 Differentiator circuit
1
the circuits discussed above. Since Z1 ¼ and
Cs
Z2 ¼ R, substitution of these values into Eq. (4.7.21) yields:

Z2
Vout ¼  Vin ¼ RCsVin
Z1
 4.7 Active-Circuit Analysis 289

Figure 4.7.11 Bandwidth-limited differentiator circuit

or in time domain
dvin
vout ¼ RC (4.7.38)
dt

As evidenced by the governing equation (Eq. (4.7.38)), the differentiator circuit reacts to
the rate of change of the input voltage. This quality is useful in control applications. When
the rate of change of the monitored parameter exceeds the allowed limit, the differentiator
circuit yields an output higher than the defined threshold, thus triggering a warning indicator
or a specified remedial action.
The sensitivity to the rate of change of the input makes the differentiator circuit very
susceptible to high-frequency noise. To combat this inherent weakness, the circuit may be
augmented as shown in Figure 4.7.11. This configuration is called a bandwidth-limited
differentiator or first-order high-pass filter.
For this circuit, the impedances Z1 and Z2 are evaluated as

1 1 þ Ri Ci s
Z1 ¼ Ri þ ¼
Ci s Ci s

and
Z2 ¼ Rf

Substituting these expressions into Eq. (4.7.21), the governing equation in the Laplace
domain is

Z2 R f Ci s
Vout ¼  Vin ¼  Vin (4.7.39)
Z1 1 þ Ri C i s

Transforming Eq. (4.7.31) to the time domain, we obtain

dvout dvin
vout þ Ri Ci ¼ Rf Ci (4.7.40)
dt dt

Another modification of the differentiator circuit is called the lead circuit and is shown in
Figure 4.7.12.
For this circuit the impedances Z1 and Z2 become
290 4 Electrical Systems

1 1 1 Ri
¼ þ ) Z1 ¼ Z 2 ¼ Rf
Z1 Ri 1=Ci s 1 þ Ri Ci s

Then, using Eq. (4.7.21), the governing equation is

derived as follows:

Z2 Rf
Vout ¼  Vin ¼  ð1 þ Ri Ci sÞVin (4.7.41) Figure 4.7.12 Lead circuit
Z1 Ri

Transforming Eq. (4.7.41) to the time domain, we obtain

Rf dvin
vout ¼  vin  Rf Ci (4.7.42)
Ri dt

Integrator–Differentiator Combination Circuits

A lead-lag or lag-lead op-amp circuit is shown in Figure 4.7.13. In this circuit, the impedances
Z1 and Z2 are represented by a parallel connection of a resistor and a capacitor.
The impedances are evaluated as follows:

1 1 1 Ri
¼ þ ) Z1 ¼
Z1 Ri 1=Ci s 1 þ Ri Ci s
1 1 Rf
Z2 ¼ þ ) Z2 ¼
Rf 1=Cf s 1 þ Rf Cf s

Substituting these expressions into Eq. (4.7.21), we Figure 4.7.13 Lead-lag

circuit
derive the governing equation for this circuit:

Z2 R f 1 þ Ri Ci s
Vout ¼  Vin ¼  Vin (4.7.43)
Z1 R i 1 þ R f Cf s

The governing equation in the time domain is then

1 dvout 1 dvin
vout þ Cf ¼  vin  Ci (4.7.44)
Rf dt Ri dt

Another interesting combination of the integrator and differentiator components – the

second-order band-pass filter – is shown in Figure 4.7.14.
This circuit is very useful in cutting off both high-frequency and low-frequency noise.
 4.7 Active-Circuit Analysis 291

Rf

Ci Cf
vin Ri vout
–

Figure 4.7.14 Second-order band-pass filter

1 1 þ Ri Ci s
The impedances Z1 and Z2 in this circuit are: Z1 ¼ Ri þ ) Z1 ¼ and
Ci s Ci s
1 1 Rf
Z2 ¼ þ ) Z2 ¼
Rf 1=Cf s 1 þ Rf Cf s
The governing equation is derived by substituting these expressions into Eq. (4.7.21) as
follows:

Z2 Rf Ci s
Vout ¼  Vin ¼    Vin (4.7.45)
Z1 1 þ Rf Cf s ð1 þ Ri Ci sÞ

In the time domain, the governing equation is a second-order differential equation:


1 Ri dvout d2 vout dvin
vout þ Cf þ Ci þ Ri C i C f ¼ Ci (4.7.46)
Rf Rf dt dt2 dt

Adder, Subtractor, and Comparator Circuits

It is possible to add, subtract, and compare signals by using the appropriate op-amp circuits.
The circuit for an adder is shown in Figure 4.7.15.
This circuit consists of two component op-amp circuits – a summing amplifier and an
inverter. The inverter was discussed above and its governing equation was derived (Eq.
(4.7.28)). To derive the relation between the inputs and the output of the summing amplifier
KCL is applied to the node A:

iR1 þ iR2 ¼ iR3 (4.7.47)

Using Ohm’s law for a current through a resistor, and remembering that the shown op-amp
inverts the sign of the input voltage, Eq. (4.7.47) becomes
vin1 vin2 vB
þ ¼ (4.7.48)
R1 R2 R3

where vB is the voltage at node B, which is an input voltage into an inverter.

292 4 Electrical Systems

R3 R

vin R1 iR R vout
1 3
– –
iR A B
1
vin R2 Inverter
2

iR
2

Summing amplifier

Figure 4.7.15 Adder circuit

R R3

vin R R1
1
vout
– –

vin Inverter R2
2

Summing amplifier

Figure 4.7.16 Subtractor circuit

Since the output of the inverter component is vout ¼ vB , the governing equation is
derived as follows:

R3 R3
vout ¼ vin1 þ vin2 (4.7.49)
R1 R2

Hence, the output voltage is a weighted sum of input voltages, where weighting factors are
defined by the values of the resistances. If all the resistors are the same – R1 ¼ R2 ¼ R3 – then
the output voltage becomes a straight sum of input voltages, so that vout ¼ vin1 þ vin2 .
In the subtractor circuit, shown in Figure 4.7.16, one of the input signals is first inverted
and then fed into a summing amplifier, while the other input is taken as is. Similarly, the
governing equation is derived as follows:
vin1 vin2 vout
 þ ¼
R1 R2 R3
(4.7.50)
R3 R3
vout ¼ vin1  vin2
R1 R2

The subtractor with equal resistances – R1 ¼ R2 ¼ R3 – is called comparator, and its

governing equation is then: vout ¼ vin1  vin2 .
 4.7 Active-Circuit Analysis 293

Non-inverting Amplifiers
All the circuits discussed in this section are inverting voltage amplifiers, meaning that the
output voltage is scaled in relation to the input voltage, and there is a change of sign between
the output and the input. In the inverting configuration, the input voltage is applied to the
inverting (−) terminal of the op-amp.
The non-inverting op-amp configuration applies the input voltage to the non-inverting (+)
terminal of the op-amp, which provides for no sign change between the output and the input,
while signal scaling still takes place.
A non-inverting amplifier circuit (multiplier) is shown in Figure 4.7.17.
The same procedure as for the inverting op-amp is used to derive the governing equation
for this circuit. As before, the grounded wire with zero electric potential is chosen as a
reference for the determination of the input and output voltages, vin and vout respectively:

vin ¼ ein  0 ¼ ein

(4.7.51)
vout ¼ eout  0 ¼ eout

Applying KCL to the node A, we obtain

i1 ¼ i 2 þ i 3 (4.7.52)

Recalling that an op-amp draws negligible current, i3 ¼ ia ≈ 0, the currents through the
resistors Ri and Rf are found to be equal:

i 1 ¼ i2 (4.7.53)

Ohm’s law derives the expressions for these currents:

0  ea  ea 
i1 ¼ ¼
Ri Ri
(4.7.54)
ea   eout ea   vout
i2 ¼ ¼
Rf Rf

Figure 4.7.17 Non-inverting voltage amplifier

294 4 Electrical Systems

Substituting them into Eq. (4.7.53) yields

ea  ea   vout
 ¼ (4.7.55)
Ri Rf

Recalling that the electric potentials on the op-amp terminals are the same, ea þ ≈ ea  (Eq.
(4.7.13)), and since ea þ ¼ ein ¼ vin , we obtain
vin vin  vout
 ¼ (4.7.56)
Ri Rf

which yields the sought governing equation:


Rf
vout ¼ 1þ vin (4.7.57)
Ri

This equation illustrates the concept of the non-inverting op-amp circuit operation – the
amplification of the input voltage by a factor that is always greater than one. There is no sign
change between the input and output voltages.
Comparing Eq. (4.7.57) with Eq. (4.7.27) demonstrates that the signal amplification
provided by a non-inverting circuit is always greater than that yielded by the similar inverting
circuit.
In addition to the discussed multiplier circuit, a non-inverting op-amp can be used in the
integrator and differentiator configurations, summing and subtracting circuits, and many
others.
Like their inverting counterparts, non-inverting op-amp circuits are used in control
systems, analog-to-digital converters, and various analog devices, including analog
computers.

Non-inverting Op-Amps as Isolation Amplifiers

With the acquired familiarity with various op-amp circuits, let us revisit the isolation
amplifier, presented in Section 4.7.1.
An isolation amplifier with a unity gain, Ga ¼ 1, is
a non-inverting op-amp with a short-circuited feed-
back, as shown in Figure 4.7.18. It may also be called
voltage follower or unity-gain buffer.
This circuit is commonly used for isolating circuits
from each other in various electronic devices. The
derivation of its governing equation is straightforward.
Figure 4.7.18 Non-inverting
A grounded wire is chosen as a reference for op-amp as an isolation amplifier
the determination of the input and output voltages,
 4.7 Active-Circuit Analysis 295

vin and vout respectively, as expressed by Eq. (4.7.51). Since the input voltage is connected
directly to the amplifier input terminal,

ein ¼ ea þ (4.7.58)

Similarly,

eout ¼ ea  (4.7.59)

Then, recalling that an op-amp draws negligible current, from Eq. (4.7.13) we obtain the
governing equation for a voltage follower:

vout ¼ vin (4.7.60)

This result can be also derived using Eq. (4.7.57). Since the feedback line in a voltage-
follower configuration is just a wire, its resistance equals one, Rf ¼ 1. The resistor Ri is
replaced by an open circuit, as seen in Figure 4.7.18, hence, Ri ¼ ∞. Substituting these values
into Eq. (4.7.57), we obtain
 
Rf 1
vout ¼ 1 þ vin ¼ 1 þ vin ¼ vin
Ri ∞

Note that a voltage follower does not provide any signal amplification; its input voltage is
passed unchanged to a circuit connected to the output terminals of this amplifier.

Converters
In control systems applications, DC signals are customarily used for the representation of
various physical measurements. In such instrumentation circuits, the measuring device acts
as a voltage source, and the controller is a load. They are typically connected in series, which
keeps the current flowing through the load exactly the same as that leaving the source. Since
the parallel connection of these two elements does not allow for equal voltages because of
resistive wire losses, a current signal is generally preferable to a voltage signal. Moreover,
current-sensing instruments are less susceptible to noise due to their low impedance, which
adds to their attractiveness.
A non-inverting amplifier circuit that is a voltage-to-current converter or trans-conductance
amplifier is shown in Figure 4.7.19.
Another integrated circuit – transistor – is added to the output wire of this circuit to
increase the range of output currents this op-amp circuit can accommodate. While the
detailed derivation of the voltage-to-current converter model is beyond the scope of this
chapter, the expression for the output current is as follows:
296 4 Electrical Systems

vin Transistor
+
iout Load
–

Figure 4.7.19 Voltage-to-current converter circuit

vin
iout ¼
R

In some cases, an electrical current is produced to

represent a physical parameter. For example, a
photodiode generates current in response to a flow
of electromagnetic radiation. Then, such a device
acts as a current source, and it may be desirable to
convert this input current to an output voltage. This
is done with an inverting op-amp circuit – a current-
to-voltage converter or trans-resistance amplifier. Figure 4.7.20 Current-to-voltage
This circuit is shown in Figure 4.7.20. converter
Its governing equation is

vout ¼ iin R

Modifications of this circuit that include a network of resistors are used when a high gain
of the converter is required.

4.7.4 Operational Amplifier Circuits with a Load

The analysis of active circuits is often aimed at understanding the evolution of the
current and voltage at the load rather than delving into processes that occur within
the op-amp component. In such a case, the op-amp circuit is represented by its
transfer function (or governing equation), and treated as an isolated voltage source.
Then, the approach to passive-circuit analysis, as described earlier in this chapter,
applies.
To illustrate this technique, consider the following example.
 4.8 Dynamic Responses via Mathematica 297

Example 4.7.2
For the active circuit shown in Figure 4.7.21, find the current passing through the load circuit.

Solution
The op-amp component is a modified differentiator circuit (lead), the governing equation of which
is represented by Eq. (4.7.41).
The governing equation for the load circuit is derived as follows:
1
I¼ Vout (1)
R þ Ls
From Eq. (4.7.41), the output voltage of the op-amp circuit is
Rf
Vout ¼  ð1 þ Ri Ci sÞVin (2)
Ri
Substituting Eq. (2) into Eq. (1), we obtain
Rf 1
I¼ ð1 þ Ri Ci sÞ Vin (3)
Ri R þ Ls
I
Hence, the transfer function of the circuit as a whole is a product of transfer functions of the
Vin
circuit components connected in series – the op-amp and the load.

Figure 4.7.21 Op-amp circuit with a load

4.8 Dynamic Responses via Mathematica

In this section, the simulation of dynamic responses of passive and active electrical circuits by
Wolfram Mathematica is illustrated in examples. Refer to Appendix C for the information
on basic programming in Wolfram Language.
298 4 Electrical Systems

Example 4.8.1
Let us revisit the circuit analyzed in Example 4.4.1 and plot the dynamic response of this system
subject to a step input.
Since the governing equations, derived by the loop method, were integral, they need to be
transformed into a differential form in order to use the Mathematica family of differential-
equation-solver functions. The linear ordinary differential equation input–output model of the
circuit in question (see Figure 4.4.1) has been derived as follows (Eq. (10) in the Example 4.4.1):
d2 vout dvout
vin ¼ R1 R2 C1 C2 2
þ ðR1 C1 þ R2 C2 þ R1 C2 Þ þ vout
dt dt
Given the following numerical values for resistances and capacitances:

R1 ¼ 5 kΩ; R2 ¼ 2:5 kΩ; C1 ¼ 5 μF; C2 ¼ 2 μF

Figure 4.8.1 Dynamic response of a circuit to the step input: (a) output voltage, (b) Mathematica code
 4.8 Dynamic Responses via Mathematica 299

find the system output – voltage across the capacitor C2 – in response to a constant input voltage
vin ¼ 10 V.
Since the governing equation is a linear second-order ordinary differential equation and the
system is subject to a step input, the closed-form solution is expected. This prompts the use of the
Mathematica DSolve function or its variant DSolveValue. The generated result is shown in
Figure 4.8.1(a), and the code is shown in Figure 4.8.1(b).
Plots of the loop currents (see Figure 4.8.2(a)) can be generated using the derived system
response. The current in loop 2 is derived from Eq. (3), Example 4.4.1, as i2 ¼ C2 v_ out . The current
in loop 1 is derived from Eq. (2), Example 4.4.1, using the above expression for
i2 : i1 ¼ C1 C2 v€out þ ðC1 þ C2 Þv_ out . These mathematical operations can be done inside the Plot
function as shown in Figure 4.8.2(b).
As seen in Figure 4.8.2(a), both loop currents are positive, which means that their directions
were assumed correctly. Additionally, as the output voltage converges to a constant value equal to
the input voltage, the currents in the circuit converge to zero.

Figure 4.8.2 Loop currents in the circuit in response to the step input voltage: (a) loop currents,
(b) Mathematica code
Example 4.8.2
Let us revisit the circuit analyzed in Example 4.4.2 and obtain plots of the dynamic responses of
this system subject to two types of input: constant (step) and periodic (sinusoidal).

Figure 4.8.3 Dynamic response of the circuit to constant and periodic input voltages: (a) response
to a step input, (b) response to a sinusoidal input, (c) Mathematica code
 4.8 Dynamic Responses via Mathematica 301

This is the same circuit as in the Example 4.4.1, but the governing equations are derived using
the node method. These equations:
8
> vin  e1 de1 e1  vout
>
< R  C1  ¼0
1 dt R 2
>
> e  vout dvout
: 1  C2 ¼0
R2 dt
are linear ordinary differential equations with no derivatives in input, so implementation in
software with the StateSpaceModel and OutputResponse functions is possible. Using the
same numerical values for resistances and capacitances as in Example 4.8.1, generate the circuit
response to two types of input voltage: a step vin ¼ 10 V, and one with a periodic component of
vin ¼ 10 þ 5 sinð100πtÞ V. The sought outputs denoted e1 and vout are voltage drops across
capacitors C1 and C2 , respectively. They are shown in Figure 4.8.3(a) and 4.8.3(b), and the
generating Mathematica code is shown in Figure 4.8.3(c).
As seen in Figure 4.8.3(a) and 4.8.3(b), the voltages across the capacitors converge to the same
constant, equal to the magnitude of input voltage for the step input, and converge to an operating
trajectory – nondecaying oscillations with constant magnitude for the sinusoidal input. Refer to
Chapter 7 for a derivation of the theoretical predictions for the observed results.

Example 4.8.3
Let us revisit the multi-source circuit analyzed in Example 4.5.2.
Given the following values of system parameters:

R ¼ 5:4 kΩ; L ¼ 2:3 mH; C ¼ 1:8 μF

and input voltages:
5uðtÞ 0 ≤ t ≤ 0:05
v1 ¼
0 t > 0:05
(pulse of the duration of 0.05 s) and v2 ¼ 4 sinð60πtÞ, derive the expression for the desired output –
current through the inductor – and obtain a plot of the system response.
Solution
This system is best implemented in software using the transfer matrix. The governing equation for
this system in terms of the current through the inductor is derived in Eq. (17) of Example 4.5.2:
d 2 iL diL dV2
LRC 2
þL þ RiL ¼ V1 þ RC
dt dt dt
Taking the Laplace transform of this equation, we derive the transfer matrix:
1 RCs
GðsÞ ¼
LRCs2 þ Ls þ R LRCs2 þ Ls þ R
Then, the system response in the Laplace domain is
302 4 Electrical Systems

Figure 4.8.4 Dynamic response of an RLC circuit with two voltage sources: (a) output current,
(b) Mathematica code
 2 . 3
5 1  e0:05s
1 RCs 6 . s 7
IL ðsÞ ¼ GðsÞRðsÞ ¼ 4 5
LRCs þ Ls þ R LRCs þ Ls þ R
2 2 4  60π  
s2 þ 3600π2

 
5 1  e0:05s 240πRCs
¼ þ
sð LRCs2 þ Ls þ RÞ ð LRCs2 þ Ls þ RÞðs2 þ 3600π2 Þ

While this expression can be inverted back to the time domain using partial fractions
expansion (refer to Chapter 2 for details), the derivation is very cumbersome and
 4.8 Dynamic Responses via Mathematica 303

time-consuming. The solution can be generated and plotted using the Mathematica
functions TransferFunctionModel and OutputResponse. The dynamic response of
the system is shown in Figure 4.8.4(a) and the generating code is shown in Figure 4.8.4(b).
High-frequency oscillations are observed at the beginning of simulation, where the influences of
pulse and periodic input combine. Then, at t ¼ 0:05 s, when the action of the pulse input ends,
these high-frequency oscillations subside, giving way to a steady nondecaying oscillatory response
due to the periodic input. A distinct jump in amplitude of response is noted at t ¼ 0:05 s due to end
of the pulse input.

Example 4.8.4
This example demonstrates signal attenuation by several active filters, described in Section 4.7.
Consider an active lag circuit (see Figure 4.7.8) that is a first-order low-pass filter, which is
expected to attenuate the signals with a frequency higher than the cutoff frequency. Defining
numeric values of system parameters as:
Rf ¼ 100 kΩ; Ri ¼ 10 kΩ; Cf ¼ 10 nF
we use Eq. (4.7.32) and the Mathematica functions TransferFunctionModel and BodePlot
to generate a Bode gain plot of this filter (see Figure 4.8.5).
As defined in Section 4.6, at the cutoff frequency ω0 the magnitude of the transfer function is

jGð jωÞjmax
jGð jωÞj ¼ pffiffiffi ffi 7:07
2

Figure 4.8.5 Bode gain plot of a first-order low-pass active filter (lag circuit)
304 4 Electrical Systems

All the signals with a frequency higher than the computed cutoff frequency (approximately 310 Hz)
will be attenuated – this filter is intended to cut off high-frequency noise in various applications. In a
loudspeaker, this filter is used on inputs on woofers and subwoofers to block high-frequency sounds
that these particular speakers cannot reproduce. Another example of the application of low-pass
filters is the tone knob on an electric guitar – here this filter is used to reduce the amount of treble in
the generated sound.
The generating Mathematica code is shown in Figure 4.8.6.
Consider a bandwidth-limited differentiator circuit (see Figure 4.7.11) that is a first-order high-
pass filter, which is expected to attenuate the signals with a frequency lower than the cutoff
frequency. Defining numeric values of system parameters as
Rf ¼ 100 kΩ; Ri ¼ 10 kΩ; Ci ¼ 0:02 μF
we use Eq. (4.7.39) and the Mathematica functions TransferFunctionModel and BodePlot
to generate a Bode gain plot of this filter (see Figure 4.8.7).
All the signals with frequency lower than the computed cutoff frequency (approximately 1.5
kHz) will be attenuated – this filter is intended to cut off low-frequency noise. One of the
applications of this filter is an audio crossover – in a loudspeaker it directs high-frequency sounds
to a tweeter while attenuating bass (low-frequency) signals that would interfere and potentially
damage the speaker.
Consider a second-order band-pass filter (see Figure 4.7.14), which is expected to pass only the
signals with a frequency within a predefined bandwidth. Defining numeric values of system
parameters as
Rf ¼ 10 kΩ; Ri ¼ 1 kΩ; Ci ¼ 20 μF; Cf ¼ 15 nF
we use Eq. (4.7.45) and the Mathematica functions TransferFunctionModel and BodePlot
to generate a Bode gain plot of this filter (see Figure 4.8.8).

Figure 4.8.6 Mathematica code for generating a Bode gain plot for a lag circuit
 4.8 Dynamic Responses via Mathematica 305

Figure 4.8.7 Bode gain plot of a first-order high-pass active filter

Figure 4.8.8 Bode plot of a second-order band-pass active filter

306 4 Electrical Systems

This band-pass filter is often referred to as narrow band-pass filter since its bandwidth, while
dependent on the resistances and the capacitances of the passive elements, is typically narrow. One
of the main applications for this filter is in communication systems for suppressing all noise and
letting the human voice come through without significant distortions.

CHAPTER SUMMARY
The main objective of this chapter is to introduce the basic physics of electrical circuits to
help the reader understand their time-dependent behavior and to learn how to model and
analyze electrical subsystems, typically encountered in a variety of engineered dynamic
systems.
Upon completion of this chapter, you should be able to:

(1) Understand the terminology of electrical circuits; refresh the knowledge of physical laws
governing common passive and active elements of electrical circuits, and understand
voltage-current relationships of these elements.
(2) Understand the concept of impedance as generalization of the electrical resistance, and
use it for modeling of electrical circuits.
(3) Derive the equivalent impedance of a circuit by using the laws of series and parallel
connection of impedances.
(4) Derive a model of a circuit by applying voltage and current divider rules and Kirchhoff’s
current and voltage laws.
(5) Use the loop method, the node method, the superposition theorem, and Norton’s and
Thevenin’s theorems for modeling and analysis of passive electrical circuits.
(6) Derive the state-space representation of electrical circuits and construct their block
diagrams.
(7) Model and analyze active electrical circuits, represented by isolation amplifiers and
operational amplifiers (op-amps).
(8) Derive transfer functions and governing equation(s) models for common op-amp cir-
cuits such as multiplier, inverter, comparator, and low- and high-pass filters.

REFERENCES
1. J. A. Svoboda and R. C. Dorf, Introduction to Electric Circuits, Wiley, 9th ed., 2013.
2. V. Hacker and C. Sumereder, Electrical Engineering: Fundamentals, De Gruyter Oldenbourg, 2020.
3. K. C. A. Smith and R. E. Alley, Electrical Circuits: An Introduction, Cambridge University Press,
1992.
 Problems 307

4. C. M. Close, D. K. Frederick, and J. C. Newell, Modeling and Analysis of Dynamic Systems, Wiley,
3rd ed., 2002.
5. N. Lobontiu, System Dynamics for Engineering Students: Concepts and Applications, Academic
Press, 2nd ed., 2017.

PROBLEMS
Section 4.2 Concept of Impedance
4.1 Two passive circuits with the indicated input and output(s) are shown in Figure P4.1.

0 0

0 0

Figure P4.1

Use the impedance method and divider rules to derive the following transfer
functions:
V0 ðsÞ I0 ðsÞ
(a) and ;
Vin ðsÞ Vin ðsÞ
V0 ðsÞ I0 ðsÞ
(b) and .
Is ðsÞ Is ðsÞ

4.2 Two passive circuits with the indicated input and output(s) are shown in Figure P4.2.

0 0

1 1
2 2

0 0

Figure P4.2

Use the impedance method and divider rules to derive the following transfer
functions:
308 4 Electrical Systems

V0 ðsÞ I0 ðsÞ
(a) and ;
Vin ðsÞ Vin ðsÞ
V0 ðsÞ I0 ðsÞ
(b) and .
Is ðsÞ Is ðsÞ

4.3 Two passive circuits with the indicated input and output(s) are shown in Figure P4.3.

Figure P4.3

Use impedance method and divider rules to derive the following transfer
functions:

V01 ðsÞ V02 ðsÞ

(a) and .
Vin ðsÞ Vin ðsÞ
V01 ðsÞ V02 ðsÞ
(b) and .
Is ðsÞ Vin ðsÞ

4.4 A passive circuit with the indicated input and output(s) is shown in Figure P4.4.

Figure P4.4

Use the impedance method and divider rules to derive the following transfer
functions:
I01 ðsÞ I02 ðsÞ V0 ðsÞ
; and :
Is ðsÞ Is ðsÞ Is ðsÞ
 Problems 309

4.5 A passive circuit with the indicated input and output(s) is shown in Figure P4.5.

Figure P4.5

Use the impedance method and divider rules to derive the following transfer
functions:
I01 ðsÞ I02 ðsÞ I03 ðsÞ
; ; and
Vin ðsÞ Vin ðsÞ Vin ðsÞ

4.6 A passive circuit with the indicated input and output(s) is shown in Figure P4.6.

Figure P4.6

Use the impedance method and divider rules to derive the following transfer
functions:
I01 ðsÞ I02 ðsÞ V0 ðsÞ
; ; and
Vin ðsÞ Vin ðsÞ Vin ðsÞ

4.7 The passive circuit with the indicated input and output(s) is shown in Figure P4.7.

Figure P4.7
310 4 Electrical Systems

Use the impedance method and divider rules to derive the following transfer
functions:
V01 ðsÞ V02 ðsÞ I0 ðsÞ
; ; and
Is ðsÞ Is ðsÞ Is ðsÞ

4.8 A passive circuit with the indicated input and output(s) is shown in Figure P4.8.

Figure P4.8

Use the impedance method and divider rules to derive the following transfer
functions:
V0 ðsÞ I0 ðsÞ
and
Vin ðsÞ Vin ðsÞ

4.9 Two passive circuits with the indicated input and output(s) are shown in Figure P4.9.

Figure P4.9

Use the impedance method and divider rules to derive the following transfer
functions:
V01 ðsÞ V02 ðsÞ I0 ðsÞ
(a) , , and ;
Vin ðsÞ Vin ðsÞ Vin ðsÞ
V01 ðsÞ V02 ðsÞ I0 ðsÞ
(b) , , and .
Is ðsÞ Vin ðsÞ Is ðsÞ
 Problems 311

4.10 A passive circuit with the indicated input and output(s) is shown in Figure P4.10.

Figure P4.10

Use the impedance method and divider rules to derive the following transfer
functions:
I01 ðsÞ I02 ðsÞ V0 ðsÞ
; ; and :
Vin ðsÞ Vin ðsÞ Vin ðsÞ

Section 4.4 Passive-Circuit Analysis

4.11 Use the loop method to derive the governing equations for the circuit shown in Figure
P4.8, considering the input ðvin Þ and required outputs ði0 ; v0 Þ indicated on the
schematic.
4.12 Use the loop method to derive the governing equations for the circuit shown in Figure
P4.10, considering the input ðvin Þ and required outputs ði01 ; i02 ; v0 Þ indicated on the
schematic.
4.13 Use the loop method to derive the governing equations for the circuit shown in Figure
P4.13, considering the input ðvin Þ and required outputs ði01 ; i02 ; v0 Þ indicated on the
schematic

Figure P4.13

4.14 Use the loop method to derive the governing equations for the circuit shown in Figure
P4.14, considering the input ðvin Þ and required outputs ði0 ; v0 Þ indicated on the
schematic.
312 4 Electrical Systems

Figure P4.14

4.15 Use the loop method to derive the governing equations for the circuit shown in Figure
P4.15, considering the input ðvin Þ and required outputs ði0 ; v01 ; v02 Þ indicated on the
schematic.

Figure P4.15

4.16 Use the loop method to derive the governing equations for the circuit shown in Figure
P4.16, considering the input ðvin Þ and required outputs ði01 ; i02 ; v0 Þ indicated on the
schematic.

Figure P4.16

4.17 Use the loop method to derive the governing equations for the circuit shown in Figure
P4.17, considering the inputs ðvin1 ; vin2 Þ and required outputs ði0 ; v0 Þ indicated on the
schematic.
 Problems 313

Figure P4.17

4.18 Use the loop method to derive the governing equations for the circuit shown in Figure
P4.18, considering the inputs ðvin1 ; vin2 Þ and required outputs ði0 ; v0 Þ indicated on the
schematic.

Figure P4.18

4.19 Use the loop method to derive the governing equations for the circuit shown in Figure
P4.19, considering the inputs ðvin1 ; vin2 Þ and required outputs ði01 ; i02 ; v01 ; v02 Þ indicated
on the schematic.

Figure P4.19
314 4 Electrical Systems

4.20 Use the loop method to derive the governing equations for the circuit shown in Figure
P4.20, considering the inputs ðvin1 ; vin2 Þ and required outputs ði01 ; i02 ; v0 Þ indicated on
the schematic.

Figure P4.20

4.21 Use the node method to derive the governing equations for the circuits shown in Figure
P4.2, considering the input and required outputs indicated on the schematic:
(a) input vin , required outputs i0 ; v0 ;
(b) input is , required outputs i0 ; v0 .

4.22 Use the node method to derive the governing equations for the circuits shown in Figure
P4.3, considering the input and required outputs indicated on the schematic:
(a) input vin , required outputs i0 ; v01 ; v02 ;
(b) input is , required outputs v01 ; v02 .
4.23 Use the node method to derive the governing equations for the circuit shown in Figure
P4.7, considering the input ðis Þ and required outputs ði0 ; v01 ; v02 Þ indicated on the
schematic.
4.24 Use the node method to derive the governing equations for the circuit shown in Figure
P4.14, considering the input ðvin Þ and required outputs ði0 ; v0 Þ indicated on the
schematic.
4.25 Use the node method to derive the governing equations for the circuit shown in Figure
P4.15, considering the input ðvin Þ and required outputs ði0 ; v01 ; v02 Þ indicated on the
schematic.
4.26 Use the node method to derive the governing equations for the circuit shown in Figure
P4.17, considering the inputs ðvin1 ; vin2 Þ and required outputs ði0 ; v0 Þ indicated on the
schematic.
 Problems 315

4.27 Use the node method to derive the governing equations for the circuit shown in Figure
P4.18, considering the inputs ðvin1 ; vin2 Þ and required outputs ði0 ; v0 Þ indicated on the
schematic.
4.28 Use the node method to derive the governing equations for the circuit, shown in Figure
P4.19, considering the inputs ðvin1 ; vin2 Þ and required outputs ði01 ; i02 ; v01 ; v02 Þ indi-
cated on the schematic.
4.29 Use the node method to derive the governing equations for the circuit shown in Figure
P4.20, considering the inputs ðvin1 ; vin2 Þ and required outputs ði01 ; i02 ; v0 Þ indicated on
the schematic.
4.30 Use the node method to derive the governing equations for the circuit shown in Figure
P4.30, considering the input ðis Þ and required outputs ði01 ; i02 ; v0 Þ indicated on the
schematic.

Figure P4.30

For problems 4.31 through 4.35 use Thevenin’s theorem to find the current through the
load and voltage drop across the load.
Recall the basic procedure for solving a circuit using Thevenin’s theorem:

1. Remove the load resistor RLoad (open-circuit it, creating two new output ter-
minals; VT is the voltage drop between these terminals).
2. Find RT by shorting all voltage sources and by open-circuiting all the current
sources, then finding the equivalent resistance of the augmented circuit.
3. Find VT by the usual circuit analysis methods (node and loop methods).
4. Find the current flowing through the load resistor RLoad and the voltage drop
across it VLoad.

4.31 For the schematic shown in Figure P4.31 use Thevenin’s theorem to find the current
through the load RLoad and the voltage drop across the load RLoad .
316 4 Electrical Systems

Figure P4.31

4.32 For the schematic shown in Figure P4.32 use Thevenin’s theorem to find the current
through the load RLoad and the voltage drop across the load RLoad .
To derive the numerical solution, use the following values for resistances and input
voltage:

R1 ¼ 7 Ω; R2 ¼ 3 Ω; R3 ¼ 2 Ω; R4 ¼ 1 Ω; vin ¼ 15 V; RLoad ¼ 2:18 Ω

Figure P4.32

4.33 For the schematic shown in Figure P4.33 use Thevenin’s theorem to find the current
through the load RLoad and the voltage drop across the load RLoad .

Figure P4.33

4.34 For the schematic shown in Figure P4.34 use Thevenin’s theorem to find the current
through the load RLoad and the voltage drop across the load RLoad .
 Problems 317

Figure P4.34

4.35 For the schematic shown in Figure P4.35 use Thevenin’s theorem to find the current
through the load RLoad and the voltage drop across the load RLoad .
To derive the numerical solution, use the following values for resistances and inputs:

R1 ¼ 8 Ω; R2 ¼ 2 Ω; R3 ¼ 4 Ω; iS ¼ 5 A; vin ¼ 10 V; RLoad ¼ 1:4 Ω

Figure P4.35

For problems 4.36 through 4.40 use Norton’s theorem to find the current through the
load and voltage drop across the load.
Recall the basic procedure for solving a circuit using Norton’s theorem:
1. Remove the load resistor RLoad (open-circuit it, creating two new output terminals).
2. Find RN by shorting all voltage sources and by open-circuiting all the current
sources, then finding the equivalent resistance of the augmented circuit.
3. Find IN by placing a shorting link on the output terminals, created in step 1, and the
usual circuit analysis methods (node and loop methods); IN is the current in the
shorting link.
4. Find the current flowing through the load resistor RLoad and the voltage drop across
it VLoad.
4.36 For the schematic shown in Figure P4.31 use Norton’s theorem to find the current
through the load RLoad and the voltage drop across the load RLoad .
318 4 Electrical Systems

4.37 For the schematic shown in Figure P4.33 use Norton’s theorem to find the current
through the load RLoad and the voltage drop across the load RLoad .
4.38 For the schematic shown in Figure P4.38 use Norton’s theorem to find the current
through the load RLoad and the voltage drop across the load RLoad .
To derive the numerical solution, use the following values for resistances and inputs:

R1 ¼ 8 Ω; R2 ¼ 12 Ω; R3 ¼ 5 Ω; iS ¼ 6 A; vin ¼ 7 V; RLoad ¼ 23:2 Ω

Figure P4.38

4.39 For the schematic shown in Figure P4.39 use Norton’s theorem to find the current
through the load RLoad and the voltage drop across the load RLoad .

Figure P4.39

4.40 For the schematic shown in Figure P4.40 use Norton’s theorem to find the current
through the load RLoad and the voltage drop across the load RLoad .
To derive the numerical solution, use the following values for resistances and inputs:

R1 ¼ 4 Ω; R2 ¼ 6 Ω; R3 ¼ 10 Ω; R4 ¼ 2 Ω; iS ¼ 1 A; vin ¼ 45 V; RLoad ¼ 3:6 Ω

Figure P4.40
 Problems 319

Section 4.5 State-Space Representations and Block Diagrams

4.41 Derive the state-space representation for the circuit shown in Figure P4.13,
considering the input ðvin Þ and required outputs ði01 ; i02 ; v0 Þ indicated on the
schematic.
4.42 Derive the state-space representation for the circuit shown in Figure P4.19, considering
the inputs ðvin1 ; vin2 Þ and required outputs ði01 ; i02 ; v01 ; v02 Þ indicated on the
schematic.
4.43 Derive the state-space representation for the circuit shown in Figure P4.20, considering
the inputs ðvin1 ; vin2 Þ and required outputs ði01 ; i02 ; v0 Þ indicated on the schematic.
4.44 Derive the state-space representation for the circuit shown in Figure P4.44, considering
the input ðis Þ and required outputs ði01 ; i02 Þ indicated on the schematic.

Figure P4.44

4.45 Derive the state-space representation for the circuit shown in Figure P4.45, considering
the input ðis Þ and required outputs ði01 ; i02 Þ indicated on the schematic.

Figure P4.45
320 4 Electrical Systems

4.46 Derive the state-space representation for the circuit shown in Figure P4.46, considering
the inputs ðvin ; is Þ and required outputs ði01 ; i02 ; v0 Þ indicated on the schematic.

Figure P4.46

4.47 Derive the state-space representation for the circuit shown in Figure P4.47, considering
the inputs ðvin ; is Þ and required output ði0 Þ indicated on the schematic.

Figure P4.47

4.48 Derive the state-space representation for the circuit, shown in Figure P4.48,
considering the inputs ðvin ; is Þ and required outputs ði0 ; v0 Þ indicated on the
schematic.

Figure P4.48
 Problems 321

4.49 Derive the state-space representation for the circuit shown in Figure P4.49,
considering the inputs ðvin1 ; vin2 Þ and required outputs ði01 ; i02 Þ indicated on the
schematic.

Figure P4.49

4.50 Derive the state-space representation for the circuit shown in Figure P4.50, considering
the inputs ðvin ; is Þ and required outputs ði01 ; i02 Þ indicated on the schematic.

Figure P4.50

4.51 Draw a block diagram for the circuit in Problem 4.11, considering the input ðvin Þ and
required outputs ði0 ; v0 Þ indicated on the schematic.
4.52 Draw a block diagram for the circuit in Problem 4.17, considering that the required
outputs are the current through the inductor and the voltage drop across the capacitor.
4.53 Draw a block diagram for the circuit in Problem 4.18, considering that the required
outputs are the voltage drops across the resistors R1 and R2 .
4.54 Draw a block diagram for the circuit in Problem 4.28, considering that the required
outputs are currents through the capacitor and through the resistor R2 , and the
potential of the node on top of the resistor R1 .
4.55 Draw a block diagram for the circuit in Problem 4.45. Use the node method to derive
the governing equations, considering that the required output is the current through
the capacitor C1 .
322 4 Electrical Systems

Section 4.6 Passive Filters

For every circuit depicted in problems 4.56 through 4.58, find the filter transfer function
Vout
GðsÞ ¼ , generate a Bode plot jGð jωÞj vs. frequency (use either Wolfram Mathematica or
Vin
MATLAB to obtain the required plots), and indicate which filter type is represented by the
given circuit.
Recall the four primary filter types:

1. Low-pass filter – a circuit that offers an easy passage to low-frequency signals ðω < ω0 Þ
while blocking high-frequency signals ðω > ω0 Þ; ω0 is a cutoff frequency in radians per
second.
2. High-pass filter – a circuit that blocks low-frequency signals (ω < ω0 ) while offering easy
passage to high-frequency signals (ω > ω0 ); ω0 is a cutoff frequency in radians per second.
3. Band-pass filter – a circuit that passes signals within a specified frequency range
ω01 < ω < ω02 while blocking signals with out-of-range frequencies; ω01 and ω02 are
the cutoff frequencies for the desired bandwidth.
4. Band-stop (notch) filter – a circuit that blocks signals within a specified frequency range
ω01 < ω < ω02 while blocking signals with out-of-range frequencies; ω01 and ω02 are the
cutoff frequencies; frequency of maximum attenuation is called notch frequency.

Both low- and high-pass filters may experience a resonance effect at the cutoff frequency,
depending on the implementation of the filter.
4.56 The circuits are shown in Figure P4.56.

Figure P4.56

The numerical data are as follows:

(a) L ¼ 5 μH; C ¼ 3 nF; R ¼ 1 kΩ;

(b) L ¼ 2 mH; C ¼ 0:5 nF; R ¼ 0:5 kΩ.
 Problems 323

4.57 The circuits are shown in Figure P4.57.

Figure P4.57

The numerical data are as follows:

(a) C ¼ 3 nF; R ¼ 1 kΩ;

(b) L ¼ 5 μH; R ¼ 1 kΩ; R1 ¼ 30 Ω; R2 ¼ 10 Ω; R3 ¼ 20 Ω; R4 ¼ 15 Ω;
(c) L ¼ 7 μH; C ¼ 1:5 nF; R ¼ 1 kΩ;
(d) C1 ¼ 1:5 nF ; C2 ¼ 1:0 nF; R1 ¼ 0:5 kΩ; R2 ¼ 1 kΩ;
(e) L ¼ 5 μH; C ¼ 3 nF; R ¼ 1 kΩ.

4.58 The circuits are shown in Figure P4.58.

The numerical data are as follows:
(a) L ¼ 5 μH; C ¼ 3 nF; R ¼ 20 Ω;
(b) C1 ¼ 1 nF; C2 ¼ 3 nF; R1 ¼ 100 Ω; R2 ¼ 1 kΩ;
(c) L1 ¼ 20 μH; L2 ¼ 3 μH; R1 ¼ 30 Ω; R2 ¼ 1 kΩ;
(d) L1 ¼ 50 mH; L2 ¼ 50 mH; C ¼ 0:1 μF; R ¼ 1 kΩ.
324 4 Electrical Systems

Figure P4.58

Section 4.7 Active-Circuit Analysis

Vout
4.59 For every active circuit shown in Figure P4.59, find the transfer function GðsÞ ¼
Vin
and derive the time-domain relationship between the input ðvin Þ and output ðvout Þ
voltages.

Figure P4.59
 Problems 325

4.60 For the differential amplifier circuit shown in Figure P4.60, derive the time-domain
relationship between the input ðvin1 ; vin2 Þ and output ðvout Þ voltages. How will this
relationship change when R1 ¼ R2 and R3 ¼ R4 ?

Figure P4.60

4.61 Derive the expression for the current i0 through the load for the Howland current
source – an active circuit, shown in Figure P4.61. Derive the balanced bridge condition –
a relationship between the resistances R1 ; R2 ; R3 ; and R4 for i0 to be independent of
the voltage drop vL across the load.

Load

Figure P4.61

4.62 The proportional–integral–derivative (PID) control – one of the commonly used linear
controls – can be implemented as an active circuit with inverting op-amps, as shown in
Figure 4.62.
326 4 Electrical Systems

Identify each op-amp circuit – the component of the shown control – and derive the
PID control law as a time-domain relationship between the input ðvin Þ and output ðvout Þ
voltages.

Figure P4.62

4.63 For the active circuit shown in Figure P4.63, derive the time-domain relationship
between the input ðvin1 ; vin2 ; vin3 ; vin4 Þ and output ðvout Þ voltages. What does this
circuit do?
How will the relationship between the circuit input and output voltages change when
R1 ¼ R2 ¼ R3 ¼ R4 ¼ Rf ?

Figure P4.63

4.64 For the active circuits shown in Figure P4.64, derive the time-domain relationships
between the input ðvin1 ; vin2 ; vin3 Þ and output ðvout Þ voltages. To show that both of these
circuits are averagers, assume that, for the circuit in Figure P4.64(a),
R1 ¼ R2 ¼ R3 ¼ R, while for the circuit in Figure P4.64(b), R1 ¼ R2 ¼ R3 ¼ 3R.
Hint: consider using Thevenin’s theorem for deriving the required expressions.
 Problems 327

Figure P4.64

4.65 The active circuits shown in Figure P4.65 represent different filters. For each of these
Vout
circuits, find the filter transfer function GðsÞ ¼ , generate a Bode plot jGð jωÞj vs.
Vin
frequency (use either Wolfram Mathematica or MATLAB to obtain the required
plots), and indicate which filter type (low-pass, high-pass, band-pass, or band-stop) is
represented by the given circuit.

Figure P4.65
328 4 Electrical Systems

The numerical data are as follows:

(a) C ¼ 100 μF; R1 ¼ 10 Ω; R2 ¼ 9 Ω; R3 ¼ 1 Ω;
(b) C ¼ 100 μF; R1 ¼ 10 Ω; R2 ¼ 9 Ω; R3 ¼ 1 Ω;
(c) C ¼ 120 μF; R1 ¼ 2 Ω; R2 ¼ 9 Ω; R3 ¼ 10 Ω; R4 ¼ 1 Ω.

4.66 A non-inverting op-amp can be used as an isolation amplifier in certain imple-

mentations of the first-order low-pass and high-pass filters, with and without
amplification. For each of the circuits shown in Figure P4.66 derive the filter
Vout
transfer function GðsÞ ¼ , generate a Bode plot jGð jωÞj vs. frequency (use either
Vin
Wolfram Mathematica or MATLAB to obtain the required plots), and indicate which
filter type is represented by the shown circuits.
For both circuits the numerical data are as follows: C ¼ 3 nF; R ¼ 1 kΩ.

Figure P4.66

4.67 A non-inverting op-amp circuit with dual capacitors is often used to obtain second-
order low-pass and high-pass filters, with and without amplification. For each of the
Vout
circuits shown in Figure P4.67 derive the filter transfer function GðsÞ ¼ , generate a
Vin
Bode plot jGð jωÞj vs. frequency (use either Wolfram Mathematica or MATLAB to
obtain the required plots), and indicate which filter type is represented by the shown
circuits.

The numerical data are as follows:

(a) C1 ¼ C2 ¼ 2 μF; R1 ¼ 20 Ω; R2 ¼ 5 Ω; R3 ¼ 100 Ω; R4 ¼ 100 Ω;

(b) C1 ¼ 0:5 μF; C2 ¼ 1:5 μF; R1 ¼ 20 Ω; R2 ¼ 1 Ω; R3 ¼ 100 Ω; R4 ¼ 10 Ω.
 Problems 329

Figure P4.67

4.68 A band-pass filter can be implemented with either inverting or non-inverting op-amps, as
shown in Figure P4.68. For each of the circuits shown in Figure P4.68 derive the filter
Vout
transfer function GðsÞ ¼ , and generate a Bode plot jGð jωÞj vs. frequency (use either
Vin
Wolfram Mathematica or MATLAB to obtain the required plots) to demonstrate the
bandwidths passed by these filters. Which one of them passed the wider bandwidth?
Numerical data are as follows:

(a) C1 ¼ 1:5 μF; C2 ¼ 1:2 μF; R1 ¼ 20 Ω; R2 ¼ 1 Ω; R3 ¼ 35 Ω; R4 ¼ 2 Ω;

(b) C1 ¼ 1:5 μF; C2 ¼ 1:3 μF; R1 ¼ 0:4 Ω; R2 ¼ 0:75 Ω.
330 4 Electrical Systems

Figure P4.68

4.69 For the active circuit shown in Figure P4.69, derive the time-domain relationship
between the input ðvin1 ; vin2 Þ and output ðvout Þ voltages. What does this circuit do?
How will the relationship between the circuit input and output voltages change when
R1 ¼ R 2 ¼ R 3 ?

Figure P4.69
 Problems 331

4.70 The inverting op-amp active circuit, shown in Figure P4.70, represents a second-order
Vout
low-pass filter. Derive the filter transfer function GðsÞ ¼ , and generate a Bode plot
Vin
jGð jωÞj vs. frequency (use either Wolfram Mathematica or MATLAB to obtain the
required plots), considering the following numerical values:

C1 ¼ 15 μF; C2 ¼ 10 μF; R1 ¼ 2:5 Ω; R2 ¼ 0:75 Ω; R3 ¼ 1 Ω:

Will this circuit pass an input signal with a frequency of 0.318 MHz? How about an
input signal with a frequency of 0.159 kHz?

Figure P4.70

4.71 One of the possible implementations of a band-stop filter (notch filter) is shown in
Vout
Figure P4.71. Derive the filter transfer function GðsÞ ¼ , and generate a Bode plot
Vin
jGð jωÞj vs. frequency (use either Wolfram Mathematica or MATLAB to obtain the
required plots), considering the following numerical values:

C1 ¼ 1:5 μF; C2 ¼ 100 nF; R1 ¼ 7:5 Ω; R2 ¼ 0:5 Ω; R3 ¼ 6 Ω; R4 ¼ 7:5 Ω; R ¼ 15 Ω:

Which frequency (in Hz) will be mostly attenuated by this circuit?

Figure P4.71
332 4 Electrical Systems

Section 4.8 Dynamic Responses via Mathematica

4.72 Consider the passive circuit shown in Figure P4.72, where system input is the voltage
vin , and the required outputs are the current i0 through the resistor R1 and the voltage
drop v0 across the inductor L2 .

Figure P4.72

Derive the governing equations for this circuit using the loop method and plot the
dynamic response of this system to a step input, considering the following numerical
values for the system parameters and input voltage:

L1 ¼ 2:5 H; L2 ¼ 1:2 H; R1 ¼ 15 Ω; R2 ¼ 25 Ω; vin ¼ 10uðtÞ V; simulation time 2:5 s:

Assume zero initial conditions.
Hint: this system has a closed-form solution for loop currents; hence, the analytical
differential equation solver DSolve of Mathematica is applicable.

4.73 Consider the passive circuit described in Problem 4.24. Derive the governing equations
for this circuit using the node method. Assuming zero initial conditions, plot the
dynamic response of this system to:
(a) a step input vin ¼ 5uðtÞ V,
(b) a ramp input vin ¼ 5tuðtÞ V,

considering the following numerical values for the system parameters and input
voltage:

L1 ¼ 2:5 H; L2 ¼ 1:2 H; R ¼ 15 Ω; C ¼ 0:65 F; simulation time 115 s:

(c) For the case of the step input, consider that capacitance is no longer constant, but
instead is dependent on time as follows: C ¼ 3ðt þ 1Þet=12 . How does the introduc-
tion of the time-dependent capacitance change the character of system response?
(d) For the case of the step input, consider that capacitance is no longer constant, but
instead is dependent on time and the voltage drop on the capacitance as follows:
C ¼ ðvC þ εÞet=12 , where ε ¼ 106 . How does the introduction of this capacitance
change the character of the system response?
 Problems 333

If the largest current the capacitor can safely handle is 10 A, will the circuits for case (c)
and case (d) be safe to operate? Explain your conclusion using the simulation results.
Hint: for this system, the numerical differential equation solver NDSolve of
Mathematica is applicable.

4.74 Consider the passive circuit described in Problem 4.44. Derive the state-space model of
this system, considering zero initial conditions and the following numerical values of
the system parameters: R1 ¼ 5 Ω; R2 ¼ 1:5 Ω; L ¼ 2 H; C1 ¼ 5 mF; C2 ¼ 2 mF:
Plot the dynamic response of this system to the following inputs:
(a) unit step input, simulation duration of 5 s;
(b) sinusoidal input 2 sinðtÞ, simulation duration of 25 s;
uðtÞ for t ≤ 2
(c) rectangular pulse input of a duration of 2 sec., described as iS ðtÞ ¼ .
0 for t > 2
Hint: for this system use the Mathematica functions StateSpaceModel and
OutputResponse.

4.75 Consider the circuit described in Problem 4.57(a). Plot the dynamic response of this
circuit to the sinusoidal input vin ðtÞ ¼ sinð10tÞ. If the input frequency is increased to
1000 rad/sec, would the periodic output still be expected? What if the input frequency is
increased to 109 rad/sec?
5 Thermal and Fluid Systems

Contents
5.1 Fundamental Principles of Thermal Systems 335
5.2 Basic Thermal Elements 339
5.3 Dynamic Modeling of Thermal Systems 346
5.4 Fundamental Principles of Fluid Systems 353
5.5 Basic Elements of Liquid-Level Systems 356
5.6 Dynamic Modeling of Liquid-Level Systems 363
5.7 Fluid Inertance 368
5.8 The Bernoulli Equation 370
5.9 Pneumatic Systems 374
5.10 Dynamic Response via MATLAB and Simulink 378
Chapter Summary 384
References 384
Problems 384

Thermal and fluid systems are found in a variety of engineering applications, including
machines, devices, automobiles, equipment, buildings, and industrial processes. Thermal
and fluid systems are in the areas of thermodynamics, heat transfer, and fluid dynamics, each
of which is the subject of a complete text, and for which there is a vast literature. The
properties of thermal and fluid systems, such as temperature, pressure, and flow, are spatially
distributed in nature, and nonlinearities usually arise in their characterization. Because of
this, distributed-parameter models governed by nonlinear partial differential equations are
required for an in-depth study of these systems. However, analytical solutions for nonlinear
partial differential equations are only limited to a few simple cases. Therefore, approxima-
tions have to be made either in modeling or during solution.
Very often, a thermal or fluid system operates with its variables changing in a small
neighborhood of a specific operating point. By using incremental variables, a linear or
linearized model of the dynamic system can be developed. Also, under certain conditions,

334
 5.1 Fundamental Principles of Thermal Systems 335

the properties of thermal and fluid components can be lumped at a point, resulting in
lumped-parameter models. This lumping process is similar to that in the modeling of
mechanical systems by using lumped elements, such as point masses, springs, and dampers.
Linear lumped-parameter models are governed by linear ordinary differential equations,
which can be systematically solved by well-developed analytical and numerical techniques.
Indeed, linear lumped-parameter models have been used in many thermal and fluid problems
and satisfactory results can be obtained.
In this chapter, we shall cover the basic concepts and important aspects of physics
regarding thermal and fluid systems. Due to the limitations of the mathematical tools used
in this undergraduate-level textbook, linear lumped-parameter models are mainly covered
for certain problems. The chapter is divided into three parts:

• modeling of thermal systems (Sections 5.1 to 5.3)

• modeling of fluid systems (Sections 5.4 to 5.9)
• numerical simulations of system responses by MATLAB and Simulink (Section 5.10)
As in dealing with mechanical systems, three keys are adopted in the modeling of thermal
and fluid systems:

• fundamental principles
• basic elements
• ways of analysis
which shall be detailed in the subsequent sections. Here, for the first time, free-body diagrams
as a way of analysis for thermal and fluid systems are introduced. The concept of the three-
key modeling technique is useful for general thermal and fluid systems, regardless of the level
of complexity of the model to be developed.
Once the governing differential equations of a lumped-parameter model are derived, they
can be converted to different forms of model representation for analysis and solution, such as
transfer function formulations, block diagrams, and state-space representations. Transfer
function formulations and state-space representations have been introduced to mechanical
systems; see Sections 3.2 and 3.7. For details on block diagrams and the use of Simulink, refer
to Sections 6.2 and 6.6. If these sections have not been covered in class or self-learning, the
reader may skip the relevant contents in this chapter and revisit them later after mastering
these topics.
In this text, the International System of Units (SI) is used. Refer to Tables A1 and A4 in
Appendix A for commonly used quantities for thermal and systems, and Table A5 for unit
conversion between the SI and the US customary system.

5.1 Fundamental Principles of Thermal Systems

A thermal system is one in which thermal energy is transferred and stored. Examples of
thermal systems include heaters, air conditioners, refrigerators, ovens, and cooling systems
336 5 Thermal and Fluid Systems

for automobiles and buildings. Thermal energy is also known as heat or heat energy.
Thermal energy transfer takes place due to temperature differences, as heat flows from an
object (location) of the higher temperature to an object (location) of the lower temperature.
The storage of heat energy in an object depends on the substance and the temperature of the
object.
Accordingly, the processes of heat transfer (thermal energy transfer) and heat storage
(thermal energy storage) are described by the following two important variables:

T – temperature, in kelvins (K) or degrees in Celsius (°C), with °C = K – 273.15

qh – heat flow rate, in joules per second (J/s) or watts (W), with 1 W = 1 J/s

Here, temperature is a measure of the potential (heat energy) of an object and heat
flow describes the heat energy transfer between two objects or locations. These two
variables are analogous to electric potential (e) and current (i) in an electric circuit; see
Section 1.4.

5.1.1 Three Ways of Heat Transfer

Heat energy can be transferred in three ways: conduction, convection, and radiation, which
are illustrated in Figure 5.1.1.

Conduction
Heat transfer by conduction occurs when two objects (locations) of different temperatures
are in physical contact, which leads to diffusion of heat through a substance (either solid
or fluid). By Fourier’s law of heat conduction, steady-state heat conduction in a one-
dimensional slab or a flat plate can be described as follows:

dT
qh ¼ kA (5.1.1)
dx

where qh is the heat flow rate; k is the thermal conductivity of the material contained in the
body, in W=ðm  KÞ or W=ðm°CÞ; A is the cross-sectional area normal to the heat flow
direction; and x is a spatial coordinate with its positive direction in that of the heat
flow. The negative sign in the previous equation indicates that there is a temperature drop
in the direction of heat flow. For a lumped-parameter model of heat conduction through a
slab, as shown in Figure 5.1.1(a), Eq. (5.1.1) is approximated by

ΔT kA
qh ≈  kA ¼ ðT1  T2 Þ (5.1.2)
Δx L

where L is the thickness of the slab in the direction of the heat flow; T1 > T2 is assumed to be
in line with the direction of qh .
 5.1 Fundamental Principles of Thermal Systems 337

A qh
qh qh Solid Fluid T2
Ts flow T1
Tf
qh
T1 Surface
T2
area A
L

(a) (b) (c)

Figure 5.1.1 Heat transfer in three ways: (a) conduction, T1 > T2 ; (b) convection, Ts > T∞ ; and
(c) radiation, T1 > T2

Convection
Heat transfer by convection involves the flow of thermal energy through the movement of a
fluid (liquid or gas). Heat convection occurs within a fluid where warmer areas of the fluid
rise to cooler areas of the fluid, as seen in water boiling in a pot. Heat convection also takes
place when a moving fluid is in touch with the surface of a solid body whose temperature is
different from that of the fluid, as seen in a heat exchanger or a fan blowing cool air through
an object. For a fluid flowing over the surface of a solid body, as shown in Figure 5.1.1(b), the
heat flow rate is given by Newton’s law of cooling,

qh ¼ hA ðTs  Tf Þ (5.1.3)
 
where h is a convective coefficient, with units of W=ðm2  KÞ and W= m2 °C ; A is the surface
area involved in heat transfer; Ts is the temperature of the solid surface; and Tf is the
temperature of the fluid at some distance away from the surface.

Radiation
Heat transfer by radiation occurs through electromagnetic waves. Unlike conduction and
convection, radiation transfers heat from one body to another without any physical contact
in between. In other words, radiation can take place in a vacuum. Examples of radiation
include heating by solar radiation and heating by heat lamps. According to the Stefan–
Boltzmann law, the heat flow rate by radiation between two separated bodies of temperat-
ures T1 and T2 , as shown in Figure 5.1.1(c), is given by

qh ¼ σ FE FA AðT14  T24 Þ (5.1.4)

where σ is the Stefan–Boltzmann constant with the value σ ¼ 5:6704  1098 kg=ðs3  K4 Þ, FE
is effective emissivity, FA is a shape factor, and A is the surface area of heat transfer.
338 5 Thermal and Fluid Systems

As can be seen from the previous discussion, conduction problems and some convection
problems can be described by linear equations. Radiation problems in general have the
nonlinearity as shown in Eq. (5.1.4), which makes it difficult in modeling and solution. In this
chapter, we shall mainly consider linear or linearized problems of heat transfer.

5.1.2 Heat Energy

For an object of mass m and specific heat c, its heat energy E is given by

E ¼ mc ðT  T0 Þ (5.1.5)

where T0 is an arbitrarily chosen reference temperature. It is assumed that the temperature

throughout the object is the same, without spatial distribution. (Refer to Section 5.2.3 for the
justification of this uniform temperature assumption.) Note that the absolute value of E is
not very meaningful. It is the change in heat energy that affects the dynamics of a thermal
system, which is similar to the potential energy of a mass m under gravity, such as mgðh  h0 Þ.
Specific heat, which is also known as specific heat capacity, is the amount of heat per unit
mass required to raise the temperature of an object by one unit. In the SI, specific heat has
units of J/(kg∙K) or J/(kg∙°C). The value of the specific heat for an object depends on the
thermodynamic process involved. Specific heat is usually measured either at constant volume
in an isochoric process, in which c is denoted by cV , or at constant pressure in an isobaric
process, in which c is denoted by cP . For an incompressible fluid or solid whose volume is
unchanged, c ¼ cV . For a constant-pressure process, as seen in the boiling of water to steam
or in the freezing of water to ice, c ¼ cP . Either way, Eq. (5.1.5) is valid for the calculation of
heat energy.

5.1.3 Conservation of Energy

All thermal systems operate under the law of conservation of energy. For a closed thermal
system with a well-defined boundary, the first law of thermodynamics, which is a version of
the law of conservation of energy, applies

ΔU ¼ Q  W (5.1.6)

where ΔU denotes the change in the energy of the system, Q represents the heat energy
supplied to the system, and W is the work done by the system on its surroundings. The
change in the energy is expressed by

ΔU ¼ ΔE þ ΔΠME (5.1.7)

where ΔE is the change in the heat energy (also called the internal energy), and ΔΠME is the
change in the mechanical energy of the system (a sum of kinetic energy and potential energy).
 5.2 Basic Thermal Elements 339

Consider a thermal system with negligible change in mechanical energy (ΔΠME ¼ 0Þ.
Assume that no work is done by the system on its surroundings (W = 0). The heat energy
supplied to the system during time interval Δt can be written as

ΔE ¼ ½qin ðtÞ  qout ðtÞΔt (5.1.8)

where qin is the rate of heat flow into the system and qout is the rate of heat flow out of the
system. Thus, the law of conservation of energy for the system reads

dE
¼ qin  qout (5.1.9)
dt

or in integral form
ðt
E ¼ ðqin  qout Þ dt (5.1.10)
t0

where t0 is an initial time.

5.2 Basic Thermal Elements

A lumped-parameter model of thermal systems has two types of basic elements: thermal
capacitance and thermal resistance. As shown in Section 1.4, these thermal elements are
analogous to capacitors and resistors in an electric circuit or springs and dampers in a mechan-
ical system. With the law of conservation of energy, the use of thermal capacitance and thermal
resistance leads to the establishment of dynamic models for thermal systems (Section 5.3).

5.2.1 Thermal Capacitance

The thermal capacitance of an object is the ability of the material of the object to store heat
energy. The thermal capacitance C of an object is defined as the rate of change of the heat
energy stored in the object with respect to the object’s temperature; that is,

dE
C¼ (5.2.1)
dT

where E is the stored heat energy as described by Eq. (5.1.5), and T is the temperature of the
object. The SI units of thermal capacitance are J/K or J/°C.
If the specific heat c of an object is not a function of temperature, the thermal capacitance
of the object by Eqs. (5.1.5) and (5.2.1) is given by

C ¼ mc ¼ ρVc (5.2.2)
340 5 Thermal and Fluid Systems

q in
T, C q out

Figure 5.2.1 A body with input and output heat flow rates

where ρ and V are the density and volume of the mass m.

Consider a body Ω of temperature T and thermal capacitance C. Assume a negligible
change in the mechanical energy in the body and zero work by the body to its surroundings.
By Eq. (5.2.1),

dE dE dT dT
¼ ¼C
dt dT dt dt

It follows from the law of conservation of energy as presented in Eq. (5.1.9) that the
governing equation of the body is obtained as follows:

dT
C ¼ qin  qout (5.2.3)
dt

Here, qin  qout is the net heat flow rate into the body, as shown in Figure 5.2.1. So, with given
input and output heat flow rates, the time-dependent temperature of the body can be
determined by solving the differential equation (5.2.3).
Comparison of Eq. (5.2.3) and Eq. (4.1.16) shows that the temperature T of a thermal
capacitance is analogous to the voltage v (potential drop) of an electrical capacitor.
Furthermore, because Eq. (5.2.3) is a first-order differential equation, the temperatures of
the thermal capacitances can be conveniently selected as state variables in the development
of a state-space model of thermal systems, as shown in Section 5.3.

5.2.2 Thermal Resistance

Thermal resistance is about a flow of heat or heat transfer through a body, and it is a
property of the material of the body that resists heat transfer. As a result of thermal
resistance, the temperature drops as heat flows through a body. This temperature drop in
a thermal-resistance element plays a similar role as a potential drop (voltage) in an electrical
resistor. In other words, thermal resistance is analogous to electrical resistance.
The thermal resistance in heat transfer is defined as

dT
R¼ (5.2.4)
dqh
 5.2 Basic Thermal Elements 341

which in the SI has units of K/W (or K  s/J) and °C=W (or °C  s=J). As presented in Section
5.1.1, there are three ways of heat transfer: conduction, convection, and radiation. For a
linear relationship between the temperature and heat flow rate, as seen in conduction models
and some convection models, Newton’s law of cooling applies:

1
qh ¼ ΔT (5.2.5)
R

where the thermal resistance is a constant for any given temperature, and ΔT is the tempera-
ture difference.
If the relationship between the temperature and heat flow rate is nonlinear, as seen in
radiation, the corresponding thermal resistance can be determined by a linearized model. To
this end, we assume that the temperature varies in a small neighborhood of a specific
temperature T . By the Taylor series,
1
qh ≈ ΔT (5.2.6)
R

with
1
R ¼ (5.2.7)
ðdqh =dT ÞT¼T

Thermal resistances in the three forms of heat transfer are obtained in sequel. For heat
conduction, as shown in Figure 5.1.1(a), Eqs. (5.1.2) and (5.2.4) give

L
R¼ (5.2.8)
kA

For heat convection between a fluid and a solid body, as shown in Figure 5.1.1(b), Eqs.
(5.1.3) and (5.2.4) yield

1
R¼ (5.2.9)
hA

For heat transfer by radiation, as shown in Figure 5.1.1(c), the relationship (5.1.4) between
the heat flow rate and the temperatures is nonlinear. Assume that the temperature T2 of the
radiated body or position is constant. Consider a small variation of the temperature of the
radiating body; that is, T1 ¼ T þ ΔT. With the linearized model given by Eq. (5.2.6) and
through the use of Eqs. (5.1.4) and (5.2.7), the thermal resistance is obtained as follows:

1
R ¼ (5.2.10)
4σ FE FA AT3

Comparison of Eq. (4.1.8) and Eq. (5.2.5) reveals that the heat flow rate qh in a thermal
system is analogous to the current i in an electrical system. Thus, in developing lumped-
parameter models, combinations of several thermal resistances can be replaced by a single
342 5 Thermal and Fluid Systems

equivalent thermal resistance, like combinations of resistors in electrical systems. For two
thermal resistances connected in series, as shown in Figure 5.2.2(a), the equations

1 1
qh ¼ ðT1  Tc Þ; qh ¼ ðTc  T2 Þ
R1 Rc

where Tc is the temperature at the interface between the thermal resistances, can be
replaced by
1
qh ¼ ðT1  T2 Þ (5.2.11)
Req

where the equivalent thermal resistance is given by

Req ¼ R1 þ R2 (5.2.12)
For two thermal resistances in parallel combination, as shown in Figure 5.2.2(b), the
equations
1 1
qh1 ¼ ðT1  T2 Þ; qh2 ¼ ðT1  T2 Þ; qh ¼ qh1 þ qh2
R1 R2

can be reduced to Eq. (5.2.11), with the equivalent thermal resistance given by
1 1 1
¼ þ (5.2.13)
Req R1 R2

For complicated thermal systems, series combinations and parallel combinations can be
further combined, as in the modeling of electrical systems.
The series combination of thermal resistances can be used to model heat conduction
through a multilayer composite slab. The parallel combination of thermal resistances can
be used to treat heat conduction through a thermal system consisting of multiple parts, as
shown in the following example.

Figure 5.2.2 Combinations of thermal resistances: (a) series combination; (b) parallel combination
 5.2 Basic Thermal Elements 343

Example 5.2.1
Figure 5.2.3 shows a closed and hollow cylindrical vessel, with thickness δ, length L, and outer
diameter D. The vessel is made of a material with thermal conductivity k. Let the temperatures
inside and outside the vessel be Ti and T0 , respectively. Determine the equivalent thermal resist-
ance of the entire vessel.
Thickness δ

qh qh
Ti , k D

T0

Figure 5.2.3 Heat conduction in a closed cylindrical vessel

Solution
The vessel has three parts: the side of the cylinder and the two circular ends. Denote the thermal
resistance of the cylinder side by Rc . Denote the thermal resistance of an end by Re . The equivalent
thermal resistance of the vessel is a parallel combination of three resistances, as shown in Figure
5.2.4. Therefore, the equivalent thermal resistance of the entire vessel is

1 1 1 1 Rc Re
Req ¼ þ þ ¼ (a)
Re Rc Re 2Re þ Rc
with the resistances Rc and Re yet to be determined.

Figure 5.2.4 Parallel combination of thermal resistances for the vessel

There are two cases of wall thickness of the vessel. For a vessel of thin wall (δ=D << 1 and
δ=L << 1), heat conduction through the thickness of the cylinder can be approximated by that
through a slab, as shown in Figure 5.1.1(a). Thus,
δ 4δ
Rc ¼ ; Re ¼ (b)
πDLk πD2 k
where Eq. (5.2.8) has been used. It follows that

2β δ
Req ¼ ; with β ¼ (c)
D þ 2L πDk
344 5 Thermal and Fluid Systems

If the wall of the vessel is thick, Eq. (5.2.8) is not applicable because the wall of the cylinder
cannot be approximated as a flat plate. In this case, the radial coordinate version of Fourier’s law
(5.1.1) can be used, and it is given by
dT dT
qh ¼ kA ¼ kð2πrLÞ ; ri ≤ r ≤ ro (d)
dr dr
D
where ri and ro are the inner and outer radii of the cylinder, and they are given by ri ¼  δ and
2
D
ro ¼ . Assume that the thermal conductivity k is constant through the thickness of the cylinder
2
wall. Integration of Eq. (d), namely,
ðro Tðo
dr
qh ¼ 2πLk dT
r
ri Ti

gives
2πLk
qh ¼ ðTi  To Þ (e)
lnðro =ri Þ
Thus, the thermal resistance of the cylinder is

lnðro =ri Þ 1 D
Rc ¼ ¼ ln (f)
2πLk 2πLk D  2δ
The thermal resistance at each end of the vessel is
h 4h
Re ¼ ¼ (g)
πri k πð D  2δÞ2 k
2

where the inner radius has been used. Substitution of the updated Rc and Re into Eq. (a) yields the
equivalent resistance of the thick-walled vessel.

In this section, two types of lumped-parameter elements have been introduced:

thermal capacitances and thermal resistances. A thermal capacitance is governed by
an ordinary differential equation like Eq. (5.2.3); a thermal resistance is described by
an algebraic equation like Eq. (5.2.5). Thermal capacitance is analogous to electrical
capacitance while thermal resistance is analogous to electrical resistance. Because of
this, a thermal system can be modeled and simulated through the use of an analogous
electrical network.
With the law of thermodynamics, however, there is no heat energy storage mechanism that
is directly related to the heat flow rate (though researchers have been trying to establish the
concept of thermal inductance in some specific cases of heat transfer). In general, there does
not exist a thermal element that is analogous to electrical inductance.
Besides thermal capacitance and resistance, thermal sources (heat energy inputs) are
specified to study the dynamic behavior of a thermal system. There are two typical types
of thermal resources:
 5.2 Basic Thermal Elements 345

• specified heat flow rate, which is positive when heat energy is added to the system or
negative when heat energy is taken away from the system; and
• specified temperature of a body as a heat energy input to the system.

5.2.3 The Biot Criterion for Lumped-Parameter Models

In a lumped-parameter model of thermal systems, the temperature within an object is assumed
to be uniform. Physically, heat transfer does occur within the body of the object, rendering the
temperature in the body spatially distributed. To justify the utility of the uniform temperature
assumption, the object needs to satisfy certain conditions or criteria. In many applications, such
criteria can be set up by the Biot number.
For a solid body immersed in a fluid, the Biot number Bi is a nondimensional number that
is defined by
hLc
Bi ¼ (5.2.14)
k

where h is a convective coefficient as given in Eq. (5.1.3); k is the thermal conductivity of the
material contained in the body; and Lc is a characteristic length of the body, which is usually
taken to be the ratio of the body volume to the area of heat transfer of the body. For instance,
for the slab in Figure 5.1.1(a),
AL
Lc ¼ ¼L
A
As another example, for a cylindrical solid of radius R and height H,

πR2 H 1 RH
Lc ¼ ¼
2πRH þ 2πR2 2 H þ R
Physically, the Biot number can be viewed as the ratio of the internal conductive thermal
resistance to the external conductive thermal resistance. For example, for a solid body
submerged in a fluid, the thermal resistance ratio is

ðRÞconduction 1=kA hL
¼ ¼ ¼ Bi (5.2.15)
ðRÞconvection 1=hA k

where Eqs. (5.2.8) and (5.2.9) have been used.

If the Biot number of a body is small enough, say Bi < 0:1, the temperature within the
body can be assumed to be uniform. This is because the conductive heat transfer within the
body experiences much less thermal resistance, compared to the convective heat transfer at
the boundary (wall) of the body. As a result, a uniform temperature distribution within the
body can be reached through relatively quick heat conduction. If the Biot number is larger
than 0.1, a uniform temperature may not be assumed. In this case, a set of coupled lumped-
parameter models (with smaller characteristic lengths) or even a distributed-parameter
model by partial differential equations is required.
346 5 Thermal and Fluid Systems

Example 5.2.2
Quenching is a metal heat-treatment process, which involves the rapid cooling of a heated
metal in water, oil, or air to improve certain material properties, like hardness and strength.
Consider an aluminum cube in quenching. The cube, with a side length of a = 0.06 mm, is
immersed in an oil tank for which the convective coefficient h = 350 W/(m2∙K). The thermal
conductivity k of aluminum varies from 220 W/(m2∙K) to 240 W/(m2∙K), in a wide tempera-
ture range of 200–700 K. By the Biot criterion, determine if the aluminum cube in quench-
ing can be treated as a lumped-parameter system with a uniform temperature.

Solution
a3 a
The characteristic length of the cube is Lc ¼ 2 ¼ ¼ 0.01 m. Take the median value of the
6a 6
thermal conductivity, k = 230 W/(m2∙K). By Eq. (5.2.14), the Biot number is

hLc 350  0:01

Bi ¼ ¼ ¼ 0:015
k 230
Because Bi is much less than 0.1, according to the Biot criterion, the aluminum cube can be treated
as a lumped-parameter system with a single uniform temperature.

5.3 Dynamic Modeling of Thermal Systems

So far, the first two keys in modeling of thermal systems have been presented: the
fundamental principles (Section 5.1) and basic elements (Section 5.2). In this section,
the last of the three keys in system modeling is introduced: a way of analysis, which
adopts free-body diagrams for thermal capacitances and auxiliary plots for thermal
resistances.
Free-body diagrams and auxiliary plots for thermal elements are simular to those for mass,
spring, and damping elements in mechanical systems (Section 3.2.2). A free-body diagram is one
in which the capacitance element in consideration is isolated from all other thermal elements, but
with input and output heat flow rates as interactions with its surroundings. An auxiliary
plot is a figure in which the resistance element with a temperature difference undergoes a
heat flow rate.
For a given thermal system of multiple thermal capacitances and resistances, the three-key
modeling technique takes the following steps:

(1) Draw a free-body diagram for each capacitance element, and an auxiliary plot for each
resistance element.
(2) Write a differential equation for each free-body diagram, and an algebraic equation for
each auxiliary plot. In this step, the formulas about the basic thermal elements given in
Section 5.2 are used.
 5.3 Dynamic Modeling of Thermal Systems 347

(3) By substitution, combine the differential and algebraic equations obtained in Step 2, to
produce a set of independent differential equations as a mathematical model of the
thermal system. The number of independent differential equations is equal to the
number of the capacitance elements in the system.
Once the governing equations of a thermal system are established, modeling tech-
niques, such as state-space representations, transfer function formulations, and block
diagrams can be applied. In a state-space formulation, it is convenient to select
capacitance temperatures as state variables. When constructing a block diagram, follow
Example 3.6.1 and use Table 3.6.1. Also, refer to Section 6.2.3 for the basic concepts
and rules about block diagrams.
The modeling approach described above is demonstrated in the following examples. For
simplicity, the following symbols are used:

a wall of perfect insulation material that prevents heat transfer through it

qin an input heat flow rate to a thermal capacitance

Example 5.3.1
In Figure 5.3.1, a thermal system consists of two capacitances that are separated by two resist-
ances. The system has two thermal inputs: a heat flow rate qin applied to the left capacitance by a
heater, and a temperature Ta (say, the ambient temperature) at the right end of the right resistance,
where heat is lost to the environment. Derive the differential equations governing the temperat-
ures T1 and T2 , obtain a state-space representation, and draw a block diagram for the system, with
T1 and the heat flow rate through the right resistance (R2 ) as the system outputs.

Solution
In the first step of system modeling, the free-body diagrams of the capacitances and the auxiliary
plots of the resistances are drawn in Figure 5.3.2, where q1 and q2 are the heat flow rates going
through the resistances.
In the second step, the equations of the capacitances and resistance are written down according
to the free-body diagrams and auxiliary plots:

Figure 5.3.1 The thermal system in Example 5.3.1

348 5 Thermal and Fluid Systems

qin q1 q1 q2

T 1 , C1 T2 , C2

Left resistance Right resistance

R1 q1 R2 q2

T1 T2 T2 Ta

Left capacitance Right capacitance

Figure 5.3.2 Free-body diagrams and auxiliary plots of the thermal system in Example 5.3.1

dT1
Left capacitance : C1 ¼ qin  q1 (a)
dt
dT2
Right capacitance : C2 ¼ q1  q2 (b)
dt
1
Left resistance : q1 ¼ ðT1  T2 Þ (c)
R1
1
Right resistance : q2 ¼ ðT2  Ta Þ (d)
R2
In the third step, by substituting Eqs. (c) and (d) into Eqs. (a) and (b), two independent
differential equations about T1 and T2 are obtained as follows:

dT1 1
C1 ¼ qin  ðT1  T2 Þ (e1)
dt R1

dT2 1 1
C2 ¼ ðT1  T2 Þ  ðT2  Ta Þ (e2)
dt R1 R2
For a state-space representation, take the capacitance temperatures as the state variables:
x1 ¼ T1 and x2 ¼ T2 . This, by Eqs. (e1) and (e2), leads to the following state equations

1 1 1
x_ 1 ¼  x1 þ x2 þ qin (f1)
C1 R1 R1

1 1 1 1 1
x_ 2 ¼ x1  þ x2 þ Ta (f2)
C2 R 1 R1 R2 R2

The output equations of the system are

1
y1 ¼ x1 ; y2 ¼ ðx2  Ta Þ (g)
R2
 5.3 Dynamic Modeling of Thermal Systems 349

where Eq. (d) has been used.

Now, for the construction of a block diagram, take Laplace transform of Eqs. (a) to (d) with
respect to time, yielding

Left capacitance : C1 sT^1 ðsÞ ¼ ^q in ðsÞ  ^

q 1 ðsÞ (h)
Right capacitance : C2 sT^2 ðsÞ ¼ ^
q 1 ðsÞ  ^q 2 ðsÞ (i)

1 ^ 
Left resistance : ^
q 1 ðsÞ ¼ T 1 ðsÞ  T^ 2 ðsÞ (j)
R1
1 ^ 
Right resistance : ^
q 2 ðsÞ ¼ T 2 ðsÞ  T^ a ðsÞ (k)
R2
where the over-hat (^) stands for Laplace transformation.
By Eqs. (h) to (k), the parts of the block diagram are plotted in Figure 5.3.3, where the basic
components in Table 3.6.1 have been used. Assembly of these parts gives the block diagram of the
thermal system in Figure 5.3.4.

Figure 5.3.3 Parts of the block diagram in Example 5.3.1

Figure 5.3.4 Block diagram for the thermal system in Example 5.3.1
350 5 Thermal and Fluid Systems

Example 5.3.2
The thermal system in Figure 5.3.5 consists a capacitance and an assembly of three insulation
layers (a series combination of three thermal resistances), where ki and δi are the thermal
conductivity and thickness of the i-th layer, respectively. Assume that all the layers have the
same cross-sectional area A. A heat flow rate qin is applied to the capacitance. On the right side of
the third layer, convective heat transfer takes place, with h being the convective coefficient, Ts the
surface temperature, and Tf the temperature of a cold fluid. This thermal system can be viewed as a
simplified model of a room or vessel with a double-pane window for insulation and protection.
Derive the governing differential equation of the system in terms of the capacitance temperature T
and determine the transfer function from qin to T and the transfer function from Tf to T.
δ1 δ2 δ3

Cold
qin fluid
k1 k2 k3
flow
Ts
Tf , h

1 2 3 Cross-section
T, C
A

Insulation layers

Figure 5.3.5 The thermal system in Example 5.3.2

Solution
The free-body diagram of the capacitance and the auxiliary plot of the resistance assembly are
shown in Figure 5.3.6, where qout is the heat flow rate out of the capacitance, R1 ; R2 ; and R3 are
the thermal resistances of the layers, R4 is the thermal resistance due to convection heat transfer,
and Req is the combined thermal resistance. The governing equations of the thermal elements by
the figure are derived as follows:

dT
Capacitance : C ¼ qin  qout (a)
dt
1  
Combined resistance : qout ¼ T  Tf (b)
Req
where the equivalent resistance Req , by Eqs. (5.2.8), (5.2.9), and (5.2.12), is given by

δ1 δ2 δ3 1
Req ¼ R1 þ R2 þ R3 þ R4 ¼ þ þ þ (c)
k1 A k2 A k3 A h A
 5.3 Dynamic Modeling of Thermal Systems 351

Figure 5.3.6 The thermal system in Example 5.3.2: (a) free-body diagram of the capacitance; (b) auxiliary
plot of the resistance

Substituting Eq. (b) into Eq. (a) gives the governing equation of the system

dT
Req C þ T ¼ Req qin þ Tf (d)
dt

Laplace transformation of Eq. (d) leads to

 
Req Cs þ 1 T^ðsÞ ¼ Req ^
q in ðsÞ þ T^f ðsÞ (e)
with which the transfer functions of the thermal system are obtained as follows:



T^ ðsÞ

Req T^ðsÞ

1

¼ ;
¼ (f)
^q in ðsÞ
^ Req Cs þ 1 ^

T f ðsÞ ^q ðsÞ¼0 R eq Cs þ1
T ðsÞ¼0
f in

Example 5.3.3
Consider the quenching process shown in Figure 5.3.7, where a heated metal piece (Tm ; Cm ) is
immersed in a liquid bath (Tb ; Cb ) of finite size. (Refer to Example 5.2.2 for a brief explanation of
quenching.) Heat transfer takes place between the solid and the liquid with thermal resistance Rb
and between the liquid and the ambient air of temperature Ta with thermal resistance Ra . Assume
that Tm > Tb . Derive a dynamic model of the quenching process in terms of the metal and bath
temperatures. Also, obtain the state equations of the process.
352 5 Thermal and Fluid Systems

Figure 5.3.7 Quenching in a liquid bath of finite size in Example 5.3.3

Solution
Because the bath is of finite size, its temperature is time-varying. By the free-body diagrams of the
capacitances and the auxiliary plots of the resistances shown in Figure 5.3.8, the equations for the
elements of the thermal system are obtained as follows:

Figure 5.3.8 Free-body diagrams and auxiliary plots of the quenching process

dTm
Metal : Cm ¼ q1 (a)
dt
dTb
Bath : Cb ¼ q1  q2 (b)
dt
1
Solidliquid heat transfer : q1 ¼ ðTm  Tb Þ (c)
Rb
1
Liquidair heat transfer q2 ¼ ðTb  Ta Þ (d)
Ra
By substitution, the dynamic model of the quenching process is described by the following coupled
differential equations:

dTm 1
Cm ¼  ðTm  Tb Þ
dt Rb
(e)
dTb 1 1
Cb ¼ ðTm  Tb Þ  ðTb  Ta Þ
dt Rb Ra
 5.4 Fundamental Principles of Fluid Systems 353

Now, by selecting two state variables

x1 ¼ Tm ; x2 ¼ Tb (f)
and with Eq. (e), the state equations of the process are obtained as follows:
1
x_ 1 ¼  ðx1  x2 Þ
R b Cm
 (g)
1 1 1 1
x_ 2 ¼ x1  þ x2 þ Ta
R b Cb R b Cb R a Cb R a Cb

5.4 Fundamental Principles of Fluid Systems

A fluid system is one in which fluids flow. A fluid may be either a liquid or a gas. Fluid
systems can be divided into two categories: (i) hydraulic systems in which the working fluids
are incompressible; and (ii) pneumatic systems in which the working fluids are compressible.
An incompressible fluid is a fluid whose density in a range of pressure does not change or has
some negligible change. Many liquids, including water, can be viewed as incompressible
fluids. On the other hand, gases, including air, are compressible fluids. Hydraulic systems are
seen in a variety of applications, including actuators and motors in manufacturing equip-
ment, chemical processes, power steering and shock absorbers in automobiles, and devices in
airplanes, rockets, and spaceships. Pneumatic systems are often used in actuators and
switches in machines and equipment, air brakes on buses and trucks, and industrial applica-
tions such as material handling, drilling, and sawing.
In the study of fluid systems, the following variables are often used:

V – volume, in m3
ρ – density or mass density, in kg/m3
q – volume flow rate (volume per unit time), in m3/s
qm – mass flow rate, qm ¼ ρq
h – liquid height, in m
p, P – pressure, in N/m2 (Pa) or atmospheric pressure (atm), with 1 atm = 101 325 Pa
T – temperature, in kelvins (K) or degrees in Celsius (°C), with °C = K− 273.15

In this section, for dynamic modeling of fluid systems by lumped-parameter models, two
types of elements are introduced: fluid capacitance and fluid resistance. Unlike thermal
systems, fluid systems also have fluid inertance due to the inertia of a moving fluid. In
some applications, like liquid-level systems, fluid inertance can be ignored due to the
negligible kinetic energy of a moving fluid. The effects of fluid inertance, however, can be
significant in other applications, as seen in piping systems. Fluid inertance is introduced in
Section 5.7.
354 5 Thermal and Fluid Systems

5.4.1 Conservation of Mass

For both hydraulic systems and pneumatic systems, the principle of mass conservation can
be stated as follows: the mass of a body that is closed to all the transfers of matter and energy
remains constant over time. For a body of a fluid with volume V and density ρ, the principle of
mass conservation can be expressed by
ð
dM d
¼ ρdv ¼ 0 (5.4.1)
dt dt
V

For an incompressible fluid ( ρ = constant), the conservation of mass is equivalent to the

conservation of volume; that is,
ð
dV d
¼ dv ¼ 0 (5.4.2)
dt dt
V

The principle of mass conservation lays out a foundation for the development of models of
fluid systems.

5.4.2 Continuity of Fluids

When a fluid moves in a duct or pipe, as shown in Figure 5.4.1(a), Eq. (5.4.1) can be written as

qm1 ¼ qm2 (5.4.3)

where qm1 and qm2 are the mass flow rates at any two locations 1 and 2. Note that a mass flow
rate can be expressed by

qm ¼ ρq ¼ ρAv (5.4.4)

Figure 5.4.1 Fluid moving in a one-dimensional duct: (a) single-body duct; (b) branched duct
 5.4 Fundamental Principles of Fluid Systems 355

where v is the velocity of the fluid, and A is the cross-sectional area with its normal in line with v.
We can substitute Eq. (5.4.4) into Eq. (5.4.3), to obtain

ρ1 A1 v1 ¼ ρ2 A2 v2 (5.4.5)

where ρi , Ai and vi are the density, cross-sectional area, and velocity of the fluid at location i,
respectively. If the fluid is incompressible ( ρ1 ¼ ρ2 ), Eq. (5.4.3) becomes

q1 ¼ A1 v1 ¼ A2 v2 ¼ q2 (5.4.6)

Equations (5.4.5) and (5.4.6) are known as continuity equations. By the same token, the fluid
continuity also holds for a branched duct, as shown in Figure 5.4.1(b), which is

ρ1 A1 v1 ¼ ρ2 A2 v2 þ ρ3 A3 v3 (5.4.7)

or
A1 v 1 ¼ A 2 v 2 þ A 3 v 3 (5.4.8)

for an incompressible fluid.

These continuity equations are analogous to Kirchhoff’s current law for electrical systems
(Section 4.3).

5.4.3 Equation of State for Gases

Gases are compressible fluids. The equation of state for an ideal gas, which is also known as
the ideal gas law, is written as

pV ¼ nRT ¼ mRg T (5.4.9)

where p, V, and T are the absolute pressure, volume, and temperature of the gas, respectively; n is
the number of moles of the gas; R is the universal gas constant, which is 8.314463 J/(mol·K); m is
the mass of the gas contained in V; and Rg is the specific gas constant depending on the
particular type of gas. Here, the mole is the unit of the amount of substance and one mole of
an element contains 6:023  1023 atoms. For dry air, Rg ¼ 287.06 J/(kg·K).
Consider a thermodynamic process from State 1 (beginning state) to State 2 (ending state).
Based on the ideal gas law, the following processes are commonly used in the development of
gas models, with the subscripts 1 and 2 denoting the beginning and ending states:

(1) Isobaric process (constant-pressure process) with p1 ¼ p2

By Eq. (5.4.9),
V2 T2
¼ (5.4.10)
V1 T1

indicating that addition of heat to the gas can raise the temperature and expand the volume
to exert external work.
356 5 Thermal and Fluid Systems

(2) Isochoric process (constant-volume process) with V1 ¼ V2

By Eq. (5.4.9),
p2 T2
¼ (5.4.11)
p1 T1

Because the volume is constant, no external work is done by the gas. Thus, in this process,
addition of heat to the gas can only raise the temperature and pressure.

(3) Isothermal process (constant-temperature process) with T1 ¼ T2

By Eq. (5.4.9),
p2 V1
¼ (5.4.12)
p1 V2

Because the temperature is constant, the addition of heat to the gas will not increase the
internal energy of the gas, but only does external work.
Besides the above three processes, the most general thermodynamic process, which is
called the polytropic process, obeys the relation
 n
V p
p ¼ ¼ constant (5.4.13)
m ρn

where ρ and m are the density and mass of the gas, respectively. If m is constant, this general
case reduces to an isobaric process for n = 0, an isochoric process for n = ∞, and an isothermal
process for n = 1. Furthermore, for n ¼ γ ≡ cP =cV (heat capacity ratio), Eq. (5.4.13) reduces
to an isentropic process (revisable adiabatic process), in which no heat is transferred to or
from the gas and which can be described by p1 V1γ ¼ p2 V2γ .
The equation of state (5.4.9) and the relation for polytropic process (Eq. (5.4.13)) are
useful in the derivation of fluid capacitance for pneumatic systems, as shown in Section 5.9.

5.5 Basic Elements of Liquid-Level Systems

Liquid-level systems are types of hydraulic systems that are assemblies of tanks (containers,
reservoirs, vessels), pipes, valves, pumps, and other components. Liquid-level systems have
wide applications in industrial processes, water treatment and water supply, and hydraulic
actuators. In the subsequent discussion, the working liquid in a liquid-level system is
assumed to be incompressible. The dynamic behavior of a liquid-level system can be
characterized by the volume flow rate q, liquid height h, and pressure p.
In this section, we introduce two types of lumped-parameter elements: fluid capacitance
and fluid resistance. Fluid capacitance corresponds to the gravitational potential energy of
the mass of a liquid in a storage tank; fluid resistance is the resistance encountered by a liquid
flowing through a pipe, a valve, or an orifice. Besides fluid capacitance and resistance, there is
a third type of element – fluid inertance – which is associated with the kinetic energy of the
mass of a liquid flowing in a pipe. In modeling liquid-level systems, fluid inertance is usually
 5.5 Basic Elements of Liquid-Level Systems 357

ignored due to the negligible kinetic energy of the mass of a flowing liquid, compared with the
gravitational potential energy of the liquid mass. Fluid inertance is useful in modeling other
fluid systems, as shown in Section 5.7.

5.5.1 Storage Tanks

Figure 5.5.1 shows a storage tank (open vessel) of arbitrary shape, which contains a liquid of
density ρ and height h. The mass of the liquid in the tank is

ðh
m ¼ ρV ¼ ρ AðyÞdy (5.5.1)
0

where AðyÞ is the cross-sectional area of the tank at height y, and V is the total volume of the
liquid. The absolute pressure of the liquid at the bottom of the tank is

p ¼ Pa þ ρgh (5.5.2)

where Pa is the atmospheric pressure (1:013  105 N=m2 ), and g is the gravitational acceler-
ation (9.807 m/s2). The gravitational potential energy of the liquid mass is PE ¼ ρVg hg ,
where hg is the height of the center of mass of the liquid in the y direction. Because
ð
1 h
hg ¼ yAðyÞdy, the potential energy is given by
V 0
ðh
PE ¼ ρg yAðyÞdy (5.5.3)
0
ð
1
The kinetic energy of the mass in the tank can be written as KE ¼ u2 dm, where u is the
2
V
magnitude of the velocity of dm. In modeling of liquid-level systems, the kinetic energy of the

liquid is usually neglected because it is much less than the potential energy given in Eq. (5.5.3).

Figure 5.5.1 A storage tank with arbitrary shape

358 5 Thermal and Fluid Systems

5.5.2 Fluid Capacitance

Fluid capacitance relates the change in stored mass to the change in pressure. The fluid
capacitance C of a tank is defined as


dm


C¼
(5.5.4)
dp

p¼pr

where m is the stored mass, and pr is a reference pressure, which is often taken as the pressure
at the bottom of the tank, pr ¼ pa þ ρgh. The SI unit of C is in kg∙m2/N. By Eqs. (5.5.1) and
(5.5.2),

dm ¼ ρAðhÞdh; dp ¼ ρgdh (5.5.5)

Thus, the fluid capacitance of the tank is

dm ρAðhÞdh AðhÞ
C¼ ¼ ¼ (5.5.6)
dp ρgdh g

Equation (5.5.6) shows that the fluid capacitance of a tank depends on the cross-sectional
area. For a tank of constant cross-section, its capacitance C is constant for any liquid
height h.
Equation (5.5.6) also indicates that fluid capacitance is independent of the fluid
density ρ. Because of this, another commonly used definition of fluid capacitance is
defined as follows:

dV
CV ¼ (5.5.7)
dh

which relates the change in the volume V of the stored mass to the change in the liquid height h.
Here, CV is in m2. Because dV ¼ AðyÞdh,

AðhÞdh
CV ¼ ¼ AðhÞ (5.5.8)
dh

Therefore,
CV ¼ Cg (5.5.9)

For clarity, in our discussion, CV is called the volume capacitance and C is called the mass
capacitance or simply capacitance.

Example 5.5.1
Determine the mass capacitance C and volume capacitance CV of the tank with the conical
frustum shape shown in Figure 5.5.2.
 5.5 Basic Elements of Liquid-Level Systems 359

r1
H A(h)

Figure 5.5.2 Conical-frustum-shaped tank in Example 5.5.1

Solution
The cross-sectional area of the tank at y = h (measured from the tank bottom) is AðhÞ ¼ πr21 , where
r1 is the radius of the liquid surface, and it is in the relation

R  r r1  r
¼
H h
This leads to

Rr
r1 ¼ hþr
H
By Eqs. (5.5.6) and (5.5.8),
 2  2
π Rr Rr
C¼ h þ r ; CV ¼ π hþr
g H H

5.5.3 Conservation of Mass

For a storage tank, the principle of mass conservation applies. In Figure 5.5.3, a tank
of liquid has volume inflow rate qin and volume outflow rate qout , in m3/s. By the principle
of mass conservation, Eq. (5.4.1), the stored mass m in the tank is governed by

dm
¼ ρqin  ρqout (5.5.10)
dt
dm dm dp
where ρ is the density of the liquid. Because ¼ , Eq. (5.5.10) is written as
dt dp dt
dp
C ¼ ρqin  ρqout (5.5.11)
dt

where Eq. (5.5.4) has been used. Furthermore, with dm ¼ ρAðhÞdh, Eq. (5.5.10) is reduced to
360 5 Thermal and Fluid Systems

Figure 5.5.3 A storage tank with inflow and outflow

dh
CV ¼ qin  qout (5.5.12)
dt

where the volume capacitance CV ¼ AðhÞ, as given in Eq. (5.5.8). Equation (5.5.12) can be
directly obtained through an examination of the change in the volume of stored liquid, which
is a special case of Eq. (5.4.2).

5.5.4 Fluid Resistance

As a liquid flows through a pipe or a valve or an orifice, a drop in the pressure of the liquid
occurs. This drop in pressure is due to the resistance that is experienced by the liquid. Fluid
resistance R can be defined as


dp


R¼
(5.5.13)
dqm

qm ¼q m

where q m is a reference mass flow rate, and R is in Pa∙s/kg or 1/(m∙s) in the SI. By the
resistance definition and a Taylor series about q m , it can be shown that

1
qm ¼ Δp (5.5.14)
R

where Δp is the change in pressure, and qm is the mass flow rate through a pipe/valve/
orifice. See Figure 5.5.4 for the symbol for fluid resistance, where Δp ¼ p1  p2 > 0. Fluid
resistance is analogous to electrical resistance, which occurs when a current goes through a
wire.
 5.5 Basic Elements of Liquid-Level Systems 361

R R
p1 p2 p1 p2 p2
p1
qm qm
R qm

Pipe Valve Orifice

R
Symbol: p1 p2

qm

Figure 5.5.4 Symbol of fluid resistance

In determining fluid resistance, the motion of a fluid needs to be considered. There are two
types of fluid motion: laminar flow and turbulent flow. For a laminar flow through a resistance
(pipe, valve, or orifice), which is “smooth” with a relatively low Reynolds number, the
pressure–mass flow rate relation is given by

p ¼ β qm (5.5.15)

where β is a constant that depends on the geometry and friction of the resistance and the
properties of the fluid. According to the definition in Eq. (5.5.13), the resistance for the
laminate flow is


dp


R¼ ¼ β (5.5.16)
dqm
qm ¼q m

Because Eq. (5.5.15) is linear, the fluid resistance R is the same at any value of q m .
For a turbulent flow through a resistance, which is “rough” with a relatively high
Reynolds number, the pressure–mass flow rate relation can be described by

p ¼ β q2m (5.5.17)

where the constant β depends on the geometry and friction of the resistance and the
properties of the fluid. Because Eq. (5.5.17) is nonlinear, the fluid resistance varies as
the mass flow rate changes. Consider a reference operating point ðq m ; pÞ, with p ¼ β q 2m . In
a small neighborhood of the operating point, the fluid resistance, with Eqs. (5.5.13) and
(5.5.17), is obtained as

dp

2p
R¼ ¼ 2β q m ¼ (5.5.18)
dqm
qm ¼ q m qm
362 5 Thermal and Fluid Systems

In the literature, the fluid resistance is also defined in terms of the volume flow rate


dh


RV ¼
(5.5.19)
dq

q¼q

where h is the liquid height, q is the volume flow rate, q is a reference flow rate, and RV in the
SI is in s/m2. Because dp ¼ ρgdh and dqm ¼ ρdq
dp ρgdh dh
¼ ¼g (5.5.20)
dqm ρdq dq

which yields
R ¼ gRV (5.5.21)

With a volume fluid resistance RV , Eq. (5.5.14) is rewritten as

1 1
qm ¼ ρq ¼ ρgΔh ¼ ρ Δh (5.5.22)
gRV RV

where Δp ¼ ðpa þ ρgh1 Þ  ðpa þ ρgh2 Þ ¼ ρgΔh has been used. It follows that
1
q¼ Δh (5.5.23)
RV
Therefore, Eqs. (5.5.14) and (5.5.23) have the same format.
Comparison of Eq. (4.1.8) with Eq. (5.5.14) or Eq. (5.5.23) indicates that the mass flow rate
qm or volume flow rate q in a liquid-level system is analogous to the current i in an electrical
system. Thus, combinations of several fluid resistances can be replaced by a single equivalent
fluid resistance, such as combinations of resistors in electrical systems. For instance, for two
fluid resistances (R1 and R2 ) connected in series, the equivalent fluid resistance is

Req ¼ R1 þ R2 (5.5.24)

and for two resistances in parallel connection, the equivalent fluid resistance is
1 1 1
¼ þ (5.5.25)
Req R1 R2

5.5.5 Orifice Resistance Formulas

An orifice can be a hole (opening) in the side of a tank, a nuzzle, or a short pipe segment with
a sudden change of cross-sectional area. A flow through an orifice experiences fluid resist-
ance, which is also described by Eq. (5.5.14), with R ¼ Ro , where Ro is the orifice resistance.
By Torricelli’s principle, the orifice resistance can be estimated by
 5.6 Dynamic Modeling of Liquid-Level Systems 363

1
Ro ¼ (5.5.26)
2ρCd2 A2o

where ρ is the mass density of the fluid, Ao is the orifice area, and Cd is the nondimensional
discharge coefficient, which is in the range 0 < Cd ≤ 1. The discharge coefficient is the ratio of
the actual mass flow rate to the theoretical one, which characterizes a head loss due to
friction. The value of Cd for water is around 0.6.

5.5.6 Sources
There are two types of energy sources that are required to operate liquid-level systems:

• flow source – a specified mass flow rate qm or volume flow rate q, which is positive when
liquid is added to a tank or negative when liquid is taken away from a tank
• pressure source – a specified increase in pressure

The flow source and pressure source to a liquid-level system correspond to the current source
and voltage source to an electric circuit.
In many liquid-level systems, a pressure source is realized by a pump (say, a centrifugal
pump). As the pump is driven by a motor at a constant speed, a pressure difference between
the inflow and outflow of the pump is generated. The pressure difference can be adjusted by
changing the rotation speed of the motor. Figure 5.5.5 shows the symbol and input–output
relationship of a pump, where qm is the mass flow rate going through the pump, and
ΔP ¼ P1  P2 is the pressure difference between the inflow and outflow of the pump.

5.6 Dynamic Modeling of Liquid-Level Systems

As mentioned previously, the modeling of a dynamic system relies on three important keys:
fundamental principle(s), basic elements, and a way of analysis. For liquid-level systems, the
fundamental principle is the law of conservation of mass, as given in Section 5.4.1 and
restated in Eq. (5.5.10) or Eq. (5.5.12); the basic elements, namely fluid capacitance and fluid
resistance, were derived in Section 5.5.

Figure 5.5.5 Symbol and input–output relation of a pump

364 5 Thermal and Fluid Systems

The last of the three keys in system modeling is a way of analysis, in which free-body
diagrams for fluid capacitances and auxiliary plots for fluid resistances are drawn. This is
similar to that for thermal systems (Section 5.3). A free-body diagram is one in which the
capacitance element is isolated from all other elements, but with input and output flow rates
as interactions with its surroundings. An auxiliary plot is one in which the resistance element
with a pressure difference undergoes a flow rate.
Therefore, the three-key modeling of a liquid-level system takes the following steps.

(1) Draw a free-body diagram for each fluid capacitance (storage tank), and an auxiliary
plot for each fluid resistance (pipe or valve or orifice).
(2) Write a differential equation for each free-body diagram, and an algebraic equa-
tion for each auxiliary plot. In this step, the element formulas given in Section 5.5
are used.
(3) By substitution, the differential and algebraic equations obtained in Step 2 are
converted to a set of independent differential equations as the dynamic model of
the system. The number of independent differential equations is equal to the
number of the capacitances in the system.

Once the governing equations of a liquid-level system are derived, state-space representa-
tions, transfer function formulations, and block diagrams (refer to Sections 3.2, 3.6, 3.7, and
6.2) can be established. In a state-space representation, it is convenient to select the liquid
height h of a storage tank as a state variable.

Example 5.6.1
A tank, with nonconstant cross-sectional area A(h), is connected to a valve at the bottom; see
Figure 5.6.1, where qin is a volume inflow rate, qout is the volume outflow rate from the valve, and
Pa is the atmospheric pressure. Derive a dynamic model of the liquid-level system, and based on
the model, obtain a state equation.

Figure 5.6.1 A tank of nonconstant cross-sectional area

 5.6 Dynamic Modeling of Liquid-Level Systems 365

Solution
The free-body diagram of the tank and the auxiliary plot of the valve are shown in Figure 5.6.2,
where q1 is the volume flow rate from the bottom of the tank to the valve. From the figure, the
governing equations of the fluid elements are written as follows:
For the tank,
dh
AðhÞ ¼ qin  q1 (a)
dt
where Eqs. (5.5.8) and (5.5.11) have been used.
For the valve:
1
ρqout ¼
ðP1  Pa Þ (b)
R
where P1 is the pressure at the bottom of the tank given by

P1 ¼ Pa þ ρgh (c)
Substituting Eq. (c) into Eq. (b) gives
g
qout ¼ h (d)
R
Equation (d) implies that the outflow rate qout is dependent upon the liquid height h and it cannot
be arbitrarily specified. By the continuity equation (5.4.6), qout ¼ q1 . It follows that the governing
equation of the liquid-level system is
dh g
AðhÞ ¼ qin  h (e)
dt R

Now select a state variable as x = h. From Eq. (e), a nonlinear state equation is obtained as follows:

Figure 5.6.2 Free-body diagram and auxiliary plot of the liquid-level system in Example 5.6.1
366 5 Thermal and Fluid Systems

1 h g i
x_ ¼ qin  x (f)
AðxÞ R

Free-body diagrams for fluid capacitances are especially useful in dealing with a system
with multiple capacitances (tanks). See the following example.

Example 5.6.2
A liquid-level system of two cylindrical tanks with constant cross-sectional areas is shown Figure
5.6.3, where the tanks are connected by Pipe 1, Tank 1 is connected to a pump (pressure source ΔP)
by a valve, Tank 2 has a specified volume inflow rate qin , and Pa is the atmospheric pressure. The
fluid resistances in the figure are all related to mass flow rates, as defined by Eq. (5.5.13). Let the
density of the fluid be ρ. Derive the governing differential equations of the system in terms of the
liquid heights h1 and h2 . Also, obtain a state-space representation with ΔP and qin as the inputs,
and the liquid height of Tank 1 and the volume outflow rate of Tank 2 through Pipe 2 as the
outputs.

Solution
The free-body diagrams of the tanks and the auxiliary plots of the pump, valve, and pipes are
shown in Figure 5.6.4, where qv , q1 , and qout are the volume flow rates. By the diagrams and plots,
the governing equations of the fluid elements are obtained as follows:
Tank 1:

dh1
A1 ¼ qv  q1 (a)
dt
Tank 2:
dh2
A2 ¼ qin  q1  qout (b)
dt

Figure 5.6.3 A two-tank liquid-level system

 5.6 Dynamic Modeling of Liquid-Level Systems 367

Figure 5.6.4 The free-body diagrams and auxiliary plots for the system in Example 5.6.2

Valve:
1
ρqv ¼ ðPa þ ΔP  P1a Þ (c)
Rp
Pipe 1:
1
ρq1 ¼ ðP1b  P2a Þ (d)
R1
Pipe 2:
1
ρqout ¼ ðP2a  Pa Þ (e)
R2
The pressures at the tank bottoms are
P1a ¼ P1b ¼ Pa þ ρgh1
(f)
P2a ¼ P2b ¼ Pa þ ρgh2

by which
ΔP g g g
qv ¼  h1 ; q1 ¼ ðh1  h2 Þ; qout ¼ h2 (g)
ρRp Rp R1 R2
Substitute Eq. (g) into Eqs. (a) and (b), to obtain the coupled differential equations of the liquid-
level system as follows:
368 5 Thermal and Fluid Systems


dh1 ΔP 1 1 g
A1 ¼ g þ h1 þ h2
dt ρRp Rp R1 R1
 (h)
dh2 g 1 1
A2 ¼ qin  h1 þ g  h2
dt R1 R1 R2

For a state-space representation, choose two state variables:

x1 ¼ h1 ; x2 ¼ h2 (i)
The state equations are obtained from Eq. (h) as follows:

dx1 g 1 1 g 1
¼ þ x1 þ x2 þ ΔP
dt A1 Rp R1 R 1 A1 ρRp A1
 (j)
dx2 g g 1 1 1
¼ x1 þ  x2 þ qin
dt R 1 A2 A2 R 1 R 2 A2

in which ΔP and qin are the inputs. The outputs of the system are
y1 ¼ h1 ; y2 ¼ qout (k)

According to Eqs. (g) and (i), the output equations are obtained as follows:
g
y1 ¼ x1 ; y2 ¼ x2 (l)
R2

5.7 Fluid Inertance

In modeling liquid-level systems, the inertia effect of a flowing fluid is neglected
because the mass of the fluid moving in pipes is much smaller than the mass of fluid
stored in tanks. However, in some other applications, like piping systems, the inertia
effect of a moving fluid can be significant and, accordingly, it cannot be ignored. For
example, when a plumbing system is suddenly shut off, the large inertia of running
water in the pipeline is brought to a quick halt and, as a result, the so-called “water
hammer” (a loud bang and then diminishing noises) is generated through shock waves.
Because of this, air chambers are usually installed on plumbing systems to reduce to
the water hammer effect.
The inertia effect of a moving fluid can be characterized by fluid inertance. For
instance, the fluid inertance I of an incompressible fluid element moving in a tube of
a constant cross-section can be identified from the momentum equation:
 5.7 Fluid Inertance 369

Figure 5.7.1 Fluid inertance of an element moving in a uniform tube

dq
I ¼ Δp (5.7.1)
dt

where q is the volume flow rate of the fluid, and Δp is the pressure difference at the
two ends of the element. This equation is similar to the equation for an electrical inductor,
di
L ¼ Δe, where Δe is the potential difference or the voltage at the two ends of the inductor.
dt
To evaluate fluid inertance, consider an incompressible fluid element of density ρ and length L
moving in a uniform tube of cross-sectional area A, as shown in Figure 5.7.1, where v is the
velocity of the fluid, q is the volume flow rate of the fluid (q ¼ Av), and P1 and P2 are the pressures
at the two ends of the element. In the theory of fluid mechanics, the element actually is a control
volume in space. By Newton’s second law, the momentum equation for the element is

d
ðmvÞ ¼ ðP1  P2 ÞA
dt

Because m ¼ ρAL and v ¼ q=A, the previous equation is reduced to

ρL dq
¼ P1  P2 ¼ Δp (5.7.2)
A dt

It follows from Eqs. (5.7.1) and (5.7.2) that the inertance of the fluid element is

ρL
I¼ (5.7.3)
A

With the fluid inertance, the kinetic energy of the element can be expressed as follows:
1
KE ¼ I q2 (5.7.4)
2

This formula can be easily verified:

1 1  q 2 1 ρL
KE ¼ mv2 ¼ ρAL ¼ q2
2 2 A 2A
1
Note that Eq. (5.7.4) is similar to the energy format of an electrical inductor, Li2 .
2
370 5 Thermal and Fluid Systems

Figure 5.7.2 Fluid inertance of an element moving in a nonuniform uniform tube

The kinetic energy expression given in Eq. (5.7.4) is useful for derivation of the inertance of
an incompressible fluid element moving in a nonuniform pipe. Figure 5.7.2 shows a fluid
element (control volume: 0 ≤ x ≤ L) in a nonuniform tube of cross-sectional area AðxÞ. Recall
that the volume flow rate q, due to the continuity equation (5.4.6), is the same at any point of
the tube. Therefore, the kinetic energy of the element can be written as

ðL ðL  2
1 1 q 1
KE ¼ ρAv dx ¼
2
ρA dx ¼ I q2 (5.7.5)
2 2 A 2
0 0

It follows that the fluid inertance I of the element is given by

ðL
dx
I¼ρ (5.7.6)
AðxÞ
0

The significance of fluid inertance in a hydraulic system must be evaluated relative to the
rest of the system and based on the frequency bandwidth of interest. Thus, the decision about
whether or not to adopt fluid inertia in system modeling depends on the application. Liquid-
level systems and piping systems are two typical examples.

5.8 The Bernoulli Equation

Consider the steady flow of an incompressible fluid, in which the velocity of the fluid at a
particular fixed point does not change with time. Steady flow is always laminar flow, but the
reverse may not be true because laminar flow can be unsteady. Assume that for a control
volume of the fluid, there is no energy exchange with the environment (heat transfer or work)
and there is no energy storage or energy dissipation (such as friction). At any point on a
streamline in the control volume, the following equation, known as the Bernoulli equation,
holds:
 5.8 The Bernoulli Equation 371

1
p þ ρgh þ ρv2 ¼ constant (5.8.1)
2

where p, ρ, and v are the pressure, density, and velocity of the fluid at the point, respectively; g
is the gravitational acceleration; and h is the elevation of the point above a reference plane,
with the positive h-direction opposite to the direction of gravity; see Figure 5.8.1. In the
1
equation, p is known static pressure and ρv2 dynamic pressure.
2
The Bernoulli equation is a statement of conservation of energy. On the left-hand side of
Eq. (5.8.1), the first term is the pressure energy of the fluid per unit volume, the second term is
the potential energy per unit volume, and the third term is the kinetic energy per unit volume.
The pressure energy per unit volume can be explained by

F Fd W Energy
p¼ ¼ ¼ ¼
A Ad V Volume

where A is the area of the fluid on which pressure force F is applied, d is a distance in the
direction of F, W is the work done by F on d, and V is the volume of the fluid. The potential
energy per unit volume and the kinetic energy per unit volume can be similarly explained.
Hence, by the Bernoulli equation, the total energy of the fluid, which is the sum of the
pressure energy, potential energy, and kinetic energy is conserved.
The Bernoulli equation is usually seen in the following equivalent form

1 1
p1 þ ρgh1 þ ρv21 ¼ p2 þ ρgh2 þ ρv22 (5.8.2)
2 2

where the subscripts 1 and 2 refer to the quantities at Point 1 and Point 2 on a streamline,
respectively, as shown in Figure 5.8.1. This form is convenient in the solution of problems in
fluid dynamics, as well as in the modeling of fluid systems.

Control volume

v
Streamline p
Point 2
g

Point 1 h

Figure 5.8.1 Pressure, elevation, and velocity of a fluid in steady flow

372 5 Thermal and Fluid Systems

Example 5.8.1
In Figure 5.8.2, water flows in a pipe of nonconstant cross-section. Assume steady flow and
negligible friction in the pipe. The pressure and velocity of water and the cross-sectional area of
the pipe at point 1 are known as P1 ¼ 2:0  105 Pa; v1 ¼ 3 m=s; A1 ¼ 0:4 m2 . At point 2, the
elevation above Point 1 and the cross-sectional area are h ¼ 2:8 m and A2 ¼ 0:25 m2 ; respectively.
Determine the pressure and velocity of the water at Point 2.

Solution
The density of water is ρ ¼ 997 kg=m3 . Take Point 1 as the reference point for elevation: h1 ¼ 0
and h2 ¼ h. By the Bernoulli equation (5.8.2)

1 1
P1 þ ρv21 ¼ P2 þ ρgh þ ρv22 (a)
2 2
Equation (a) has two unknows: P2 and v2 . Therefore, one more equation is needed. Because water
is an incompressible fluid, the continuity equation (5.4.6) applies

A1 v 1 ¼ A2 v 2 (b)

From Eq. (b), the velocity of the fluid at Point 2 is found as

A1 0:4
v2 ¼ v1 ¼  3 ¼ 4:8 m=s (c)
A2 0:25

With the known v2 , the solution of Eq. (a) gives the pressure at point 2 as follows:
1  
P2 ¼ P1  ρgh þ ρ v21  v22
2
1  
¼ 2:0  10  997  9:807  2:8 þ 997  32  4:82
5
2
¼ 1:656  105 Pa

A2
v2
Pipe
A1 g
p2
Point 2
h
v1
p1

Point 1

Figure 5.8.2 Steady flow of water in a nonuniform pipe in Example 5.8.1

 5.8 The Bernoulli Equation 373

Modified Bernoulli Equation

Equation (5.8.2) and Example 5.8.1 assume negligible friction in the pipe. In many applica-
tions, such as a fluid flowing in a long pipeline, the friction between the fluid and the pipe
walls can be significant. Under the circumstances, the Bernoulli equation is often modified.
To this end, we define the total head H at a point on a streamline by

p 1
H¼ þ h þ v2 (5.8.3)
ρg 2g

p 1
where is called the pressure head, h the elevation head, and v2 the velocity head. Also,
ρg 2g
p
þ h is known as the piezometric head or hydraulic head. The H represents the total energy
ρg
per weight without energy loss, and it is also known as the available head. With H, the
Bernoulli equation (5.8.2) can be written in the following equivalent form

H1 ¼ H2 (5.8.4)

Equation (5.8.4) means that the total head of a fluid system without energy loss is constant.
For a pipe with energy loss due to friction, a modified Bernoulli equation is written as

H1 ¼ H2 þ hf (5.8.5)

or
p1 1 p2 1
þ h1 þ v21 ¼ þ h2 þ v22 þ hf (5.8.6)
ρg 2g ρg 2g

where hf is the head loss due to friction, characterizing the energy (pressure) loss in the pipe.
The hf is a function of the geometry and internal surface roughness of the pipe and the fluid
speed. Depending on applications, the value of hf can be just a few percent of H in some
cases, and it may be between 10% to 20% in other cases. Empirical formulas, such as
the Darcy–Weisbach equation, are commonly used to estimate the head loss in a pipeline.
Due to limited space, these formulas are not covered here.

Example 5.8.2 Due to friction, the water pipe in Figure 5.8.2 has a head loss hf ¼ 0:86 m from
Point 1 to Point 2. Let the pressure and velocity of water at Point 1 be the same as in Example
5.8.1. Determine the pressure and velocity of water at Point 2.
374 5 Thermal and Fluid Systems

Solution
By the continuity equation (5.4.6), the velocity of water at Point 2 is the same as in Example 5.5.1;
that is, v2 = 4.8 m/s. The total head at Point 1 is
p1 1
H1 ¼ þ h1 þ v21
ρg 2g
(a)
2:0  105 32
¼ þ0þ ¼ 20:995 m
997  9:807 2  9:807
By the modified Bernoulli equation (5.8.5),
p2 1
H1 ¼ H 2 þ h f ¼ þ h2 þ v22 þ hf (b)
ρg 2g
which, after algebraic manipulations, gives the pressure at Point 2 as follows:
  1
p2 ¼ ρg H1  hf  ρgh2  ρv22
2
1 (c)
¼ 997  9:807  ð20:995  0:86Þ  997  9:807  2:8  997  4:82
2
¼ 1:576  105 Pa

Comparison of the result with that of Example 5.8.1 shows that the friction in the pipe results in
a 4.8% pressure drop (or pressure head loss) at Point 2.

5.9 Pneumatic Systems

The working fluid in a pneumatic system is a gas, which is a compressible fluid. Air is most
commonly used in many such systems. Although the properties of pneumatic systems are highly
nonlinear and spatially distributed, as in modeling of hydraulic systems, lumped-parameter
models can be developed to obtain satisfactory results in many applications. In this section, we
follow the same three-key modeling technique as presented in Sections 5.5 and 5.6.

5.9.1 Fundamental Principle

A fundamental principle for pneumatic systems is the principle of mass conservation, which
has been presented in Section 5.4.1. There are two things that differentiate a pneumatic
system from a hydraulic system. First, the gas in a pneumatic system, due to its compress-
ibility, has changeable density. Second, a pneumatic system exhibits different behaviors in
different thermodynamic processes, during which temperature, pressure, volume, and mass
are in a certain relationship, such as the ideal gas law.

5.9.2 Basic Elements

Pneumatic systems in general have three types of lumped-parameter elements: pneumatic
capacitance, pneumatic resistance, and pneumatic inertance. The kinetic energy of a gas in
many applications is negligible, compared to the potential energy due to the pressure.
 5.9 Pneumatic Systems 375

Because of this, inertance is usually ignored and only capacitance and resistance are con-
sidered in system modeling of pneumatic systems.
Similarly to the definition of fluid capacitance in Section 5.5, the pneumatic capacitance C
relates the change in a stored gas mass to the change in gas pressure; that is,

dmg
C¼ (5.9.1)
dp

where mg is the stored gas mass; C has units of kg  m2 =N or kg/Pa.

Consider a container of constant volume V that stores a gas of density ρ. Because mg ¼ ρV

dðρV Þ dρ
C¼ ¼V (5.9.2)
dp dp

For an ideal gas undergoing a polytropic process as described by Eq. (5.4.13),


p dp pdρ
d n ¼ n  n nþ1 ¼ 0
ρ ρ ρ

or
dρ ρ
¼
dp np

From the ideal gas law of Eq. (5.4.9), p ¼ ρRg T, the previous equation becomes

dρ 1
¼ (5.9.3)
dp nRg T

It follows from Eqs. (5.9.2) and (5.9.3) that the pneumatic capacitance of the container is

V
C¼ (5.9.4)
nRg T

Equation (5.9.4) shows that the capacitance C of a constant-volume container can

be different in different thermodynamic processes (n and T) and for different gases
(Rg ).
Assume that a constant-volume container has a mass inflow rate qmi and a mass outflow
rate qmo . According to the principle of mass conservation, Eq. (5.4.1), the mass in the
container is governed by the differential equation

dmg dmg dp
¼ ¼ qmi  qmo
dt dp dt
376 5 Thermal and Fluid Systems

p1 R p2
Gas

qm

Figure 5.9.1 Pneumatic resistance of a valve or orifice

which, by the definition (5.9.1), gives the governing equation of the container pressure as
follows:

dp
C ¼ qmi  qmo (5.9.5)
dt

where the container pressure p is the dynamic variable. For the gas undergoing a polytropic
process, Eq. (5.9.5) can be written as

V dp
¼ qmi  qmo (5.9.6)
nRg T dt

When a gas flow passes through a valve or an orifice, it meets with resistance. As a result, a
pressure drop in the gas occurs, as shown in Figure 5.9.1, where qm is the mass flow rate, and the
pressure drop is Δp ¼ p1  p2 . The pneumatic resistance R in the figure can be expressed by

1
qm ¼ Δp (5.9.7)
R

and R is in Pa∙s/kg. Note that the relation between the mass flow rate and pressure in general
is nonlinear. Therefore, the value of R is obtained for a small neighborhood of a reference
operating point ðq m ; pÞ, which is similar to that of fluid resistance for a turbulent flow
(Section 5.5).
Besides capacitance and resistance elements, pneumatic systems require energy sources to
operate. There are two types of sources: (i) flow sources, maintaining a specified mass flow
rate qm ; and (ii) pressure sources, maintaining a specified pressure or pressure increase. These
energy sources can be realized by compressors, ventilators, gas supply tanks, and pressure
control valves.

5.9.3 Modeling of Pneumatic Systems

As in the modeling of liquid-level systems, the development of models of pneumatic systems
is based on three keys: (i) the fundamental principle (conservation of mass); (ii) basic
elements (pneumatic capacitance and resistance); and (iii) a way of analysis (free-body
diagrams of capacitances and auxiliary plots of resistances). This modeling technique is
demonstrated in the following example.
 5.9 Pneumatic Systems 377

Example 5.9.1
In Figure 5.9.2, a rigid cylinder of constant volume V is connected to a tank of compressed air by a
valve of resistance R. The air-supply tank is a source with a constant pressure Ps . Assume that the
gas-filling process in the cylinder is a polytropic process, with Ps > P. Derive a model of the
pneumatic system in terms of the cylinder pressure P.

Solution
The solution procedure is similar to that for a liquid-level system. From the free-body diagram of
the cylinder and the auxiliary plot of the valve shown in Figure 5.9.3, the governing equations of
the pneumatic elements are as follows:
V dP
Cylinder : ¼ qm (a)
nRg T dt
1
Valve : qm ¼ ðPs  PÞ (b)
R

where Eqs. (5.9.6) and (5.9.7) have been used. It follows from Eqs. (a) and (b) that the governing
differential equation of the pneumatic system is

Figure 5.9.2 Pneumatic system for Example 5.9.1

Figure 5.9.3 Free-body diagram and auxiliary plot for Example 5.9.1
378 5 Thermal and Fluid Systems

RV dP
þ P ¼ Ps (c)
nRg T dt

For pneumatic systems with multiple capacitances, model development takes steps that
are similar to those in the modeling of liquid-level systems (see Section 5.6).

5.10 Dynamic Response via MATLAB and Simulink

In this section, the simulation of dynamic responses of thermal and fluid systems by MATLAB
and Simulink is illustrated in examples. Refer to Appendix B for the use of MATLAB and
Simulink.

Example 5.10.1
Consider the two-capacitance thermal system in Figure 5.3.1, with parameter values given
as: C1 ¼ 500 J/K, C2 ¼ 300 J/K, R1 = 12 K  s/J, R2 = 800 K  s/J. Let the input heat flow rate be
qin ¼ q0 eσt , with q0 = 326 J/s and σ = 0.01/s. Let the outside temperature be Ta = 250 K. Assume
the initial temperatures as follows: T1 ð0Þ ¼ T2 ð0Þ ¼ 258 K. Plot the temperatures T1 ðtÞ and T2 ðtÞ
1
and the heat flow rate qout ðtÞ ¼ ðT2 ðtÞ  Ta Þ, for 0 ≤ t ≤ 500 s.
R2

Solution
From Example 5.3.1, the matrix state equation of the thermal system is

ẋ ¼ Ax þ Bu; xð0Þ ¼ x0 (a)

where
  
T1 q0 eσt T1 ð0Þ
x¼ ; u¼ ; x0 ¼
T2 Ta T2 ð0Þ

2 3 2 3
1 1 1
6 C1 R1 C 1 R1 7 6 C1 0 7
6 7 6 7
A¼6  7; B¼6 7 (b)
4 1 1 1 1 5 4 5
 þ 0 1
C2 R1 C2 R 1 R 2 C2 R2

The state equation (a) can be solved numerically by the fixed-step Runge–Kutta method of
order four, as presented in Section 2.8. The following commands, which are contained in a scrip
M-file and a function M-file plot the system response (T1 ; T2 ; qout ) in Figure 5.10.1. In the M-file
RKfunc5111, A and B are the matrices from Eq. (b).
 5.10 Dynamic Response via MATLAB and Simulink 379

% Input data
C1 = 500; C2 = 300; R1 = 12; R2 = 800;
q0 = 326; sgm = 0.01; Ta = 250;
T10 = 258; T20 = 258;

% Compute system response

B = [1/C1 0; 0 1/C2/R2];
t_stop = 500; npts = 5001; h = t_stop/(npts-1);
z0 = [T10; T20]; % initial state
t = linspace(0, t_stop, npts);
z = zeros(2, npts);
z(:,1) = z0;
for i = 1: npts-1
tt = t(i);
zz = z(:,i);
f1 = RKfunc5111(tt, zz, A, B, q0, sgm, Ta);
f2 = RKfunc5111(tt+h/2, zz+h/2*f1, A, B, q0, sgm, Ta);
f3 = RKfunc5111(tt+h/2, zz+h/2*f2, A, B, q0, sgm, Ta);
f4 = RKfunc5111(tt+h, zz+h*f3, A, B, q0, sgm, Ta);
z(:,i+1) = zz +h/6*(f1+2*f2+2*f3+f4);
end

% Plot system response

figure(1) % Temperatures
plot(t,z(1,:),’k’, t,z(2,:),’k–’)
xlabel(‘Time (s)’), ylabel(‘Temperature (K)’)
grid, legend(‘T_1‘, ‘T_2’)

figure(2) % Heat flow rate

qout = (z(2,:)-Ta)/R2;
plot(t,qout,’k’), grid
xlabel(‘Time (s)’), ylabel(‘Heat Flow Rate, q_o_u_t (J/s)’)

function f = RKfunc5111(t,z, A, B, q0, sgm, Ta)

u = [q0*exp(-sgm*t); Ta];
f = A*z + B*u;

The numerical solutions in this example can also be obtained by Simulink. For this, the basics of
block diagrams (Sections 6.2) and Simulink (Section 6.6 and Appendix B) are required. By Eq. (e)
of Example 5.3.1, a Simulink model is established in Figure 5.10.2, where the Simulink blocks
Clock and Fcn are used to generate the time-dependent input qin . Running a simulation on this
model produces the same plots in Figure 5.10.1.
380 5 Thermal and Fluid Systems

320
T1
310 T2

300
Temperature (K)

290

280

270

260

250
0 50 100 150 200 250 300 350 400 450 500
Time (s)
(a)
0.02

0.019

0.018
Heat Flow Rate, qout (J/s)

0.017

0.016

0.015

0.014

0.013

0.012

0.011

0.01
0 50 100 150 200 250 300 350 400 450 500
Time (s)
(b)

Figure 5.10.1 Dynamic response of the thermal system in Example 5.10.1: (a) the temperatures
of the capacitances; and (b) the output heat flow rate of the right capacitance
 5.10 Dynamic Response via MATLAB and Simulink 381

L +– 1/500 1
f(u)
q_in 8
1/C1 T1

1/12 +
–
1/R1

++ 1
1/300
8
1/C2 T2

+ 1/800
–
1/R2 q_out
250
T_a

Figure 5.10.2 A Simulink model of the thermal system in Example 5.11.1

Example 5.10.2
Consider the liquid-level system in Figure 5.6.3, with the system parameter values given by:
ρ = 1000kg/m3, g = 9.81 m/s2, A1 ¼ 3.5 m2, A2 ¼ 4 m2, Rp ¼ 230 Pa∙s/kg, R1 ¼ 300 Pa∙s/kg, R2 ¼
Pa∙s/kg. Let the pump pressure difference be Δp ¼ p0 uðtÞ, where uðtÞ is the unit step function, and
p0 = 120 kPa. The volume inflow rate of Tank 2 is qin ¼ 0.07 m3/s. Assume the initial liquid heights
h1 ð0Þ ¼ 1.5 m and h2 ð0Þ ¼ 0.3 m. Plot the liquid heights h1 ðtÞ and h2 ðtÞ of the tanks, for 0 ≤ t ≤
1200 s.

Solution
From Example 5.6.2, the matrix state equation of the liquid-level system is

ẋ ¼ Ax þ Bu; xð0Þ ¼ x0 (a)

where
  
h1 Δp h1 ð0Þ
x¼ ; u¼ ; x0 ¼
h2 qin h2 ð0Þ
and
2  3 2 3
g 1 1 g 1
6  A 1 Rp þ R1 7 6 0 7
A¼6  R1 A1 7; B ¼ 6ρRp A1 7 (b)
4 g g 1 1 5 4 1 5
  0
R1 A2 A 2 R1 R2 A2

By following Example 5.10.1, the use of the fixed-step Runge–Kutta method of order four results
in the plots of the liquid heights shown in Figure 5.10.3.
382 5 Thermal and Fluid Systems

2.5

Liquid Heights (m)

1.5

0.5 h1
h2

0
0 200 400 600 800 1000 1200
Time (s)

Figure 5.10.3 The liquid heights of the system in Example 5.10.2

Example 5.10.3
Consider the pneumatic system shown in Figure 5.9.2, where a rigid cylinder is connected to a
supply tank of dry air by a valve. Assume an isothermal (constant temperature) process. The
volume and temperature of the rigid cylinder are V = 20 m3 and T = 298 K. The valve resistance is
R = 1200 Pa∙s/kg. Let the pressure of the air-supply tank be Ps ¼ 265 kPa. Determine the pressure
PðtÞ of the cylinder, which initially is Pð0Þ = 101 kPa.

Solution
From Example 5.9.1, the pressure of the cylinder is governed by
RV dP
þ P ¼ Ps (a)
nRg T dt

where n = 1 for an isothermal process (Section 5.4.3) and Rg ¼ 287.06 J/(kg·K) for dry air. Laplace
transform of Eq. (a) gives the s-domain cylinder pressure as follows:

^ μPð0Þs þ Ps RV
PðsÞ ¼ ; with μ ¼ (b)
sðμs þ 1Þ nRg T
^ is the Laplace transform of PðtÞ. Inverse Laplace transform of PðsÞ
where PðsÞ ^ gives the solution of
Eq. (a).
By the following MATLAB commands:
 5.10 Dynamic Response via MATLAB and Simulink 383

% Input data
R = 1200; V = 20; T = 298;
Rg = 287.06; n = 1;
ps = 2.65e5; p0 = 1.01e5;

% Compute and plot response

mu = R*V/(n*Rg*T)
num = [mu*p0 ps];
den = [mu 1 0]/1000;
impulse(num, den);
xlabel(‘t’)
ylabel(‘Pressure (kPa)’)
grid, title(‘Pressure p(t) of the Rigid Cylinder’)
the cylinder pressure is plotted in Figure 5.10.4. Here, the MATLAB function impulse from
Table B8 in Appendix B has been used for the inverse Laplace transform of PðsÞ. ^ Note that the
steady-state value of the cylinder pressure is always equal to Ps . This can also be verified by the
final-value theorem:

^ μPð0Þs þ Ps
Pss ¼ lim PðtÞ ¼ lim sPðsÞ ¼ lim ¼ Ps
t→∞ s→0 s→0 μs þ 1

×108
2.8

2.6

2.4

2.2
Pressure (kPa)

1.8

1.6

1.4

1.2

1
0 0.5 1 1.5 2 2.5
t (seconds)

Figure 5.10.4 Cylinder pressure p(t) in Example 5.10.3

384 5 Thermal and Fluid Systems

CHAPTER SUMMARY
This chapter emphasizes three keys in modeling of thermal and fluid systems: funda-
mental principles, basic elements, and ways of analysis. For thermal systems, three
forms of heat transfer and the law of conservation of energy are reviewed; lumped-
parameter thermal elements are presented. For fluid systems, the principle of mass
conservation, the continuity equations, the ideal gas law, and the Bernoulli equation
are reviewed; lumped-parameter fluid elements are developed for liquid-level systems
and pneumatic systems. For both thermal and fluid systems, free-body diagrams,
which borrow the concepts from the modeling of mechanical systems, are introduced
as a way of analysis. After a mathematical model is devised, it is converted to a
transfer function formulation, block diagram and state-space representations, for
simulation and analysis. Finally, the utility of MATLAB and Simulink to compute
the response of fluid and thermal systems is illustrated.
Upon completion of this chapter, you should be able to:

(1) Understand and use the three important keys in modeling of thermal and fluid systems.
(2) Derive the governing equations of thermal systems, liquid-level systems, and pneumatic
systems, with lumped-parameter models.
(3) Establish the transfer function formulation, block diagram, and state-space representa-
tion for a given thermal or fluid system.
(4) Compute the dynamic response of modeled thermal and fluid systems by MATLAB and
Simulink.

REFERENCES
1. Y. Cengel, M. Boles, and M. Kanoglu, Thermodynamics: An Engineering Approach, 9th ed.,
McGraw-Hill Education, 2018.
2. T. L. Bergman, A. S. Lavine, F. P. Incropera, and D. P. DeWitt, Introduction to Heat Transfer, 6th
ed., Wiley, 2011.
3. R. C. Hibbeler, Fluid Mechanics, 2nd ed., Pearson, 2017.

PROBLEMS
Section 5.1 Fundamental Principles of Thermal Systems
5.1 Consider heat transfer through a single-pane glass window in Figure P5.1, where To and
Tr are the air temperatures outside and inside the room, respectively; T1 and T2 are the
temperatures on the outer and inner surfaces of the glass, respectively. The glass layer
has thickness l, area A, and thermal conductivity k. Let the heat transfer coefficient for
convection between the outside air and the outer glass surface be h1 . Let the heat transfer
 Problems 385

coefficient for convection between the inner glass surface and the room air be h2 . Assume that
the parameters of the thermal system have the following values: l = 1 cm, A = 1.8 m2, k = 0.85
W=ðm  °CÞ, h1 ¼ 12 W=ðm2  °CÞ, and h2 ¼ 10 W=ðm2  °CÞ. Assume that To = 32 °C and
Tr = 24 °C.
(a) Determine the heat flow rate qh through the window.
(b) Determine the temperatures T1 and T2 .

Outdoor Glass Room

air air
T1 T2
To Tr
qh

Figure P5.1

5.2 Consider heat transfer through a three-layer wall in Figure P5.2, where Ta1 and Ta2 are
the air temperatures near the sides of the wall; lj and kj are the thickness and thermal
conductivity of the j-th layer, for j = 1, 2, 3; and h1 and h2 are the heat transfer coefficients
for convection on the left and right surfaces of the wall, respectively. The layers have the
same cross-sectional area A. Assume that Ta1 > Ta2 .
(a) Determine the heat flow rate qh through the wall.
(b) Determine the temperature distribution through the wall.

h1 h2
k1 k2 k3
Ta1 Ta2
qh

l1 l2 l3

Figure P5.2

Section 5.2 Basic Thermal Elements

5.3 Consider a tank of 1 liter of water, with ρ = 1000 kg/m3 and cP ¼ 4:186  103 J/(kg∙°C).
Here 1 liter = 10−3 m3. A reference temperature is chosen as T0 = 0 °C.
(a) Compute the amount of heat energy stored in the water at temperature T = 20 °C.
386 5 Thermal and Fluid Systems

(b) Compute the thermal capacitance of the water.

(c) How much energy does it take to raise the water temperature from 20 °C to 90 °C?

5.4 For the three-layer wall in Figure P5.2, determine the equivalent thermal resistance Req
in the relation
1
qh ¼ ðTa1  Ta2 Þ
Req

5.5 Consider a closed and hollow rectangular container of thickness δ, outer length a, outer
width b, and outer height h. The container is thin-walled (δ << a; b; c) and is made of a
material with thermal conductivity k. Let the temperatures inside and outside the
container be Ti and To , respectively. Determine the equivalent thermal resistance of
the entire vessel.
5.6 Consider a hollow aluminum cylinder of inner diameter di , outer diameter do , height H,
and thermal conductivity k. The outer surface of the cylinder is exposed to a stream of air
with temperature To and velocity uo . On the inner surface a constant temperature Ti is
maintained. The convective heat transfer coefficient between the outer surface of the
cylinder and the stream of air is approximated by h ¼ 3:15 uo , in W=ðm2  KÞ. Consider
the parameter values: di = 0.5 m, do = 0.75 m, H = 1 m, k = 240 W=ðm  KÞ.
(a) Compute the characteristic length Lc of the aluminum cylinder.
(b) By the Biot criterion, determine the condition of the velocity uo of air under which
the cylinder can be treated by a lumped-parameter model.

Section 5.3 Dynamic Modeling of Thermal Systems

5.7 A thin-walled container of volume V is fully filled with a liquid, which has mass
density ρ and specific heat c. The container wall has a total thermal resistance
(convective and conductive) Rc and negligible thermal capacitance. The temperature
of the air surrounding the container is Ta . Let the liquid temperature in the container
be Tl .
(a) Draw a free-body diagram for the thermal capacitance about the liquid, and an
auxiliary plot for the thermal resistance about the container wall.
(b) Derive a mathematical model of the liquid temperature Tl , with Ta as the input.
(c) Let the initial liquid temperature be τ0 . Determine the liquid temperature Tl ðtÞ by
solving the differential equation obtained in part (b).
5.8 The thermal system in Figure P5.8 consists of two capacitances that are separated by two
resistances. A heat flow rate qin is applied to the left capacitance. On the right side of
resistance R2 , convective heat transfer takes place, with h being the convective coefficient
and Tf the temperature of a fluid stream.
 Problems 387

(a) Draw free-body diagrams for the thermal capacitances, and auxiliary plots for the
thermal resistances.
(b) Derive the differential equations governing T1 and T2 .

qin Fluid
Ts
Tf , h
R1 R2
T1 ,C1 T2 ,C2

Cross-section A

Figure P5.8

5.9 Consider the thermal system in Figure P5.8, with qin and Tf as the inputs.
(a) Derive the transfer function with qin as the input and the heat flow rate through the
right resistance (R2 ) as the output.
(b) Draw a block diagram for the system, with qin and Tf as the inputs and T2 as the
output.
(c) Obtain a state-space representation for the system, with qin and Tf as the inputs, and
the heat flow rate through the left resistance (R1 ) and the temperature of the right
capacitance (T2 ) as the outputs.
5.10 A simplified model of metal heat treatment is shown in Figure P5.10, where a metal part
of temperature Tm and thermal capacitance Cm is placed inside an industrial furnace
that is heated by a heater at rate qin . In the figure, T1 and C1 are the temperature and
thermal capacitance of the air inside the furnace; T2 and C2 are the temperature and
thermal capacitance of the furnace wall; and Ta is the ambient temperature. Assume
that heat is transferred by convection only, with hm and Am being the convective
coefficient and surface area between the metal part and the air inside the furnace; h1
and A1 , between the air inside the furnace and the inner wall; and h2 and A2 , between the
outer wall and ambient air.
(a) Draw free-body diagrams of the thermal capacitances, and auxiliary plots of the
thermal resistances.
(b) Derive the differential equations governing Tm , T1 , and T2 .
388 5 Thermal and Fluid Systems

Wall

qin T1 , C1

Tm , Cm
Ta

T2 , C2

Figure P5.10

5.11 Consider the thermal system in Figure P5.10, with qin and Ta as the inputs.
(a) Derive the transfer function with qin as the input and Tm as the output.
(b) Obtain a state-space representation for the system, with Tm and the heat flow rate
on the surface of the outer wall as the system outputs.
5.12 Consider the quenching process similar to that in Figure 5.3.7, with the following
modification: the volume of the liquid in the bath is large enough so that the bath
temperature Tb remains unchanged when the heated metal piece is immersed in the
bath. With the modification, the heat transfer between the liquid and the ambient air is
ignored, as can be seen from Eq. (b) of Example 5.3.3 with Cb → ∞.
(a) Derive a mathematical model for the temperature Tm of the metal piece.
(b) Consider quenching of a steel sphere of radius r. The parameters of the sphere are given
as: r = 0.02 m, ρ = 7730 kg/m3, cP = 420 J/(kg∙ °C). The convective coefficient for heat
transfer between the sphere and the liquid is h = 360 W/(m2∙°C). The temperature of
the liquid bath is 60 °C. The initial temperature of the sphere is 820 °C. By the
model obtained in part (a), plot Tm versus time and determine the time needed for
the sphere temperatures to reach 200 °C.

Section 5.4 Fundamental Principles of Fluid Systems

5.13 A cylinder–piston assembly is connected to a spring–mass system by a rod; see
Figure P5.13, where A is the cross-sectional area of the piston, and p1 and p2 are
the pressures applied to the sides of the piston by two pumps. Under the pressures, the
piston slides on the frictionless surface of the cylinder, with displacement x. Let the
combined inertia of the piston and rod be M. Assume that the spring is not deformed at
x = 0.
(a) Derive a mathematical model for the piston displacement x.
(b) For p1  p2 ¼ P0 , where P0 is a nonzero constant, determine the displacement x(t)
assuming that the piston is initially at rest.
 Problems 389

p1 p2
x

Rod k
A M

Cylinder
Piston

Figure P5.13

5.14 Consider the hydraulic system in Figure P5.14, where an external force f is
applied to the left piston of area A1 , and, under the hydraulic pressure, the right
piston of area A2 moves up to lift mass m. The mass is constrained by a spring and
damper pair (k, c). Assume that m includes the inertia of the right piston. The inertia
effect of the fluid in the hydraulic system is ignored. Assume that the inertia of the left
piston is negligible.
(a) Derive a model for the displacement x2 of the right piston.
(b) Determine the work done by the external force f in terms of the displacement x2 .

f k c

x2
x1 A1
A2

Figure P5.14

Section 5.5 Basic Elements of Liquid-Level Systems

5.15 A spherical tank is shown in Figure P5.15, where qm is an input mass flow rate, and h is
the liquid head. Let the liquid density be ρ. Let the diameter of the tank be D.
(a) Determine the mass capacitance C and volume capacitance CV for the tank.
(b) Derive a dynamic model of the height h of the tank.
390 5 Thermal and Fluid Systems

qm

Figure P5.15

5.16 Consider the cylindrical tank in Figure P5.16, where qm is the input mass flow rate, and
h is the liquid head. Let the liquid density be ρ.
(a) Determine the mass capacitance C and volume capacitance CV for the tank.
(b) Derive a dynamic model of the liquid height h of the tank.

qm

h
L

Figure P5.16

5.17 For a long tube with a laminar flow, the fluid resistance can be calculated by the
formula
32μl
R¼
ρAdh2

where l and A are the length and the area of the tube; μ and ρ are the dynamic viscosity
(in Pa∙s) and density (in kg/m3) of the fluid, respectively; and dh is the hydraulic
diameter of the tube given by dh ¼ 4A=S, with S being the perimeter of the tube area.
Here, μ and ρ are dependent on the temperature.
 Problems 391

(a) For a tube with a circular cross-section of diameter D, show that its fluid resistance is

128μl
R¼
πρD4
(b) For a tube with a square cross-section of width w, show that its fluid resistance is

32μl
R¼
ρw4
(c) Consider a circular tube, which is 1 m in length and 5.0 mm in diameter. Determine
the fluid resistance of the tube conveying the following two fluids.

Fluid 1: water at 25 °C, with μ = 8.90 104 Pa∙s and ρ = 997 kg/m3.
Fluid 2: motor oil (SAE 10 W-40) at 60 °C, with μ = 37.147 104 Pa∙s and ρ =
838 kg/m3.

Section 5.6 Dynamic Modeling of Liquid-Level Systems

5.18 Consider the two-tank liquid-level system in Figure P5.18, where qin is the input
volume flow rate (in m3/s), pa is the atmospheric pressure, the tanks have constant
cross-sectional areas (A1 ; A2 ), and the fluid flow is laminar in the valves.
(a) Draw free-body diagrams for the tanks, and auxiliary plots for the valves.
(b) Derive a model of the system with liquid heights h1 and h2 as the dynamic variables.
H2 ðsÞ Q2 ðsÞ
(c) Determine the transfer functions and .
Qin ðsÞ Qin ðsÞ

pa
qin

A1 pa
h1 R1
q1
Tank 1 Valve 1

A2 pa
h2
R2 q2

Tank 2 Valve 2

Figure P5.18
392 5 Thermal and Fluid Systems

5.19 Consider the liquid-level system in Figure P5.19, where a spherical tank of
diameter D is connected to an outlet valve (R), qin is the input volume flow rate,
and pa is the atmospheric pressure. Refer to Problem 5.15 for the fluid capacitance of
the tank.
(a) Derive a model of the system with liquid height h as the dynamic variable.
(b) Obtain a state-space model of the system, with qin as the input, and qout as the
output.

qin pa

A(h)
h pa
R
qout

Valve

Figure P5.19

5.20 A liquid-level system is shown in Figure P5.20, where two tanks of constant areas
(A1 , A2 ) are connected by a pump with pressure difference Δp and a valve; Tank 1 has a
volume inflow rate qin ; Tank 2 has a volume outflow rate qout ; and pa is the atmospheric
pressure. The fluid resistances R and R1 are related to the mass flow rate, as described
by Eq. (5.5.14). The density of the fluid is ρ.
(a) Draw free-body diagrams for the tanks, and auxiliary plots for the pump, valve,
and pipe.
(b) Derive the governing differential equations of the system with liquid heights h1 and
h2 as the dynamic variables.
(c) Obtain a state-space model with qin and Δp as the inputs, and h2 and qout as the
outputs.

pa pa
qin

A2 pa
A1
h1 h2
Δp R R1 qout

Tank 1 Pump Valve Tank 2 Pipe

Figure P5.20
 Problems 393

5.21 The liquid-level system in Figure P5.21 consists of two tanks of constant areas (A1 , A2 ),
a branched inflow pipeline with two valves (R1 and R2 ), a pipe (R3 ) connecting the
tanks, and an outflow pipe (R4 ) from tank 2. In the figure, qin and qout are the volume
inflow rate and outflow rate, respectively; the fluid resistances are related to the mass
flow rate, as described by Eq. (5.5.14); and pa is the atmospheric pressure. Let the fluid
density be ρ.
(a) Draw free-body diagrams for the tanks, and auxiliary plots for the valves and
pipes.
(b) Derive the governing differential equations of the system in terms of the liquid
heights h1 and h2 .
(c) Obtain a state-space model with qin as the input, and h2 and qout as the outputs.

qin R2

Valve R1 Valve pa

A2 pa
A1
h1 h2
R3 R4 qout

Tank 1 Pipe Tank 2 Pipe

Figure P5.21

5.22 The liquid-level system in Figure P5.22 consists a single tank of constant area A, which
has an orifice at height ho , a pump with pressure difference Δp, and a valve (R); qin is a
volume inflow rate, R is related to the mass flow rate, and pa is the atmospheric
pressure. Let the area and discharge coefficient of the orifice be Ao and Cd , respectively.
The density of the fluid is ρ. Assume that the liquid height h is always larger than ho .
(a) Draw a free-body diagram for the tank, and auxiliary plots for the pump, valve,
and orifice.
(b) Derive the governing differential equation of the system in terms of h.

qin pa

A Orifice
h
pa Δp R ho

Pump Valve Tank

Figure P5.22
394 5 Thermal and Fluid Systems

Section 5.8 The Bernoulli Equation

5.23 In Figure P5.23, a tank of constant area A has an orifice that is a distance d from the
liquid level; pa is the atmospheric pressure, and qin a volume inflow rate that is properly
adjusted to maintain the liquid head H unchanged. At the orifice, there is a head loss hf
due to friction. Given parameter values H ¼ 1 m, d = 0.6 m, and hf = 0.05 m, determine
the speed of the liquid emerging from the orifice.

pa
qin

A d
H
pa

Figure P5.23

Section 5.9 Pneumatic Systems

5.24 A rigid spherical container with a diameter of 0.5 m is filled with dry air. Compute the
pneumatic capacitance of the container in an isothermal process, with the air at an
ambient temperature of 25 °C.
5.25 Consider the pneumatic system in Figure P5.25, where two rigid cylinders of constant
volumes (V1 , V2 ) are connected by a valve of resistance R2, and the left cylinder is
connected to an air-supply tank of a constant pressure ps by a valve of resistance R1. Let
the gas-filling process in the cylinders be an isothermal process, with ps > p1 and
ps > p2 .
(a) Draw free-body diagrams of the cylinders and auxiliary plots of the valves.
(b) Derive the governing differential equations of the cylinder pressures p1 and p2 .

R2

From supply
R1
tank, ps
p1, C1 p2, C2

Cylinder 1 Cylinder 2

Figure P5.25
 Problems 395

Section 5.10 Dynamic Response via MATLAB and Simulink

5.26 Consider the thermal system in Figure P5.8. The right side of resistance R2 has a cross-
sectional area A and a convective coefficient h. The system parameters have the
following values: C1 = 9500 J/°C°C/W, C2 = 16000 J/°C, R1 = 0.12 °C/W, R2 = 0.082,
h = 115 W=ðm2  °CÞ and A = 1 m2. Let the inputs to the system be qin = 250 W and
Tf = 5 °C. The initial temperatures of the capacitances are T1 ð0Þ = 15 °C and T2 ð0Þ ¼
10 °C. Use the MATLAB functions impulse and step to plot the temperature T2 of
the right capacitance versus time.
5.27 Repeat Problem 5.26 by building a Simulink model and running simulation.
5.28 Consider the liquid-level system in Figure P5.19, with D = 2 m and R = 500 Pa∙s/kg. The
liquid density is ρ = 1000 kg/m3 and the input volume flow rate qin = 0.02 m3/s.
The initial liquid height is h(0) = 0.15 m. By obtaining a nonlinear state-space model
of the system and using the fixed-step Runge–Kutta method of order four (Section 2.8),
plot the liquid height h(t) and the output volume flow rate qout , for 0 ≤ t ≤ 600 s.
5.29 Consider liquid-level system in Figure P5.21. The system parameters have the following
values: A1 = 3 m2, A2 = 6 m2, R1 = 150 Pa∙s/kg, R2 = 200 Pa∙s/kg, R3 = 500 Pa∙s/kg, R4 =
600 Pa∙s/kg. The input volume flow rate is qin = 0.04 m3/s. The initial heights of the
tanks are h1 ð0Þ = 0.5 m and h2 ð0Þ = 1.2 m. By Simulink, determine the liquid heights h1
and h2 .
5.30 Consider the pneumatic system in Figure P5.25. The volumes of the cylinders are
V1 = 20 m3 and V2 = 40 m3. The resistances of the valves are R1 = 900 Pa∙s/kg,
R2 = 1200 Pa∙s/kg. Assume an isothermal (constant-temperature) process, with the
cylinder temperatures T1 = T2 = 298 K. The pressure of the air-supply tank is
ps ¼ 300 kPa. The initial pressures of the cylinders are p1 ð0Þ = p2 ð0Þ = 101 kPa. By
Simulink, plot the cylinder pressures p1 and p2 versus time.
6 Combined Systems and System Modeling
Techniques

Contents
6.1 Introduction to System-Level Modeling 397
6.2 System Modeling Techniques 398
6.3 Fundamentals of Electromechanical Systems 425
6.4 Direct Current Motors 433
6.5 Block Diagrams in the Time Domain 457
6.6 Modeling and Simulation by Simulink 464
6.7 Modeling and Simulation by Wolfram Mathematica 475
Chapter Summary 485
References 486
Problems 486

The main objective of this chapter is to introduce the concept of a dynamic system consisting
of multiple subsystems, and to present the reader with the typical approach to modeling and
analysis of such a combined system. A detailed discussion on combined systems is conducted
using an electromechanical system – direct current (DC) motor – as an example. More
instances of combined systems such as sensors and actuators (electromechanical, piezoelec-
tric, piezoresistive), thermomechanical devices, and electroacoustic transducers are pre-
sented in Chapter 9.
The first section of this chapter presents terminology and basic concepts of system-level
modeling for any combined dynamic system. Section 6.2 offers a detailed discussion on the
typical system modeling techniques such as transfer function formulations, state-space
representations, and block diagrams in the Laplace domain. Block diagrams in the time
domain are discussed further in Section 6.5. Sections 6.3 and 6.4 offer a brief review of the
foundations of electromagnetism and describe physical processes occurring in the electro-
mechanical systems using the example of a DC motor. The derived mathematical models of
the armature-controlled and field-controlled DC motors are implemented in software:

396
 6.1 Introduction to System-Level Modeling 397

MATLAB/Simulink (Section 6.6) and Wolfram Mathematica (Section 6.7); system behav-
iors in response to different inputs – (a) constant input and (b) periodic excitation – are
simulated and the generated results are analyzed.

6.1 Introduction to System-Level Modeling

The commonly encountered dynamic systems often comprise multiple subsystems that may
belong to different fields. For example, a popular dynamic system such as a DC motor is a
combination of two subsystems: electrical and mechanical. Since the majority of engineering
systems are controlled, they can be considered as compound systems having at least two
different components: a plant element, represented by a mechanical, electrical, fluid, or
thermal subsystem, or any combination of those; and the controller element.
The ability to model and analyze such complex dynamic systems hinges on an understand-
ing of the interactions between subsystems in addition to having a firm grasp on the behavior
of the individual system components. A typical approach to system-level modeling and
analysis is illustrated in Figure 6.1.1.
The first step, as depicted in Figure 6.1.1, involves deriving a set of governing equations for
each subsystem separately, and obtaining mathematical expressions of their coupling rela-
tions that describe interactions between these subsystems. Obviously, the subsystem’s gov-
erning equations contain only the variables of the field to which this subsystem belongs. For
example, a mechanical subsystem will be described with variables such as displacements,

Complex dynamic system

Mechanical Electrical Thermal Fluid Feedback

subsystem subsystem subsystem subsystem controllers

Synthesis

System analysis

Validation of optimal design

Figure 6.1.1 System-level modeling and analysis flowchart

398 6 Combined Systems and System Modeling Techniques

velocities, and accelerations, while an electrical subsystem is described with currents and
voltages. To the contrary, the coupling expressions involve variables of different types.
At the synthesis stage, the subsystems are joined together into a single system model. That
is done by the derivation of any of the following three representations: (a) system transfer
functions, (b) system state-space representations, or (c) system block diagrams. Since any
one of these three representations can be obtained from any other one in a relatively simple
and straightforward manner, the derivation of a single representation is usually sufficient for
subsequent system analysis.
For the linear time-invariant systems that are the subject of this book, all of the above
types of system representation are equally suitable. In general, the choice of the type of
system model depends on the nature of the compound system. For example, the applicability
of a transfer function formulation is limited by the existence of the Laplace transform of the
governing equation. As discussed in detail in Chapter 2, the Laplace transform is best used
for linear time-invariant systems that are described by a set of linear ordinary differential
equations. Thus, for a system described by a first-order differential equation such as
ð4 þ sinð3tÞÞx_ þ 5x ¼ FðtÞ, the Laplace transform does not exist, making a transfer function
formulation unavailable.
Block diagrams and state-space representations are available for both linear and nonlinear
systems, whether time-invariant or not. They perform well in modeling multi-input/multi-
output systems and are exceptionally well suited for model implementation in modeling and
simulation software such as MATLAB/Simulink by MathWorks, or Mathematica/System
Modeler by Wolfram Research. They are particularly helpful in modeling nonlinear systems
efficiently, and in facilitating their analysis.
The derivation of all three representations from a set of governing equations and the
conversion of one type of system model into another is discussed later in this section. The
synthesis step concludes system-level modeling.
System analysis includes the implementation of the synthesized system model in the
chosen software, followed by simulations. The derived set of system-governing equations
is solved to obtain a system response; the system performance in response to the previously
defined inputs is studied, and the system stability is analyzed.
The last step in system-level analysis is the validation of the optimal design. On the basis of
performance and stability data, obtained during system analysis, the necessary design
changes are made and the process is repeated until a satisfactory design is produced.

6.2 System Modeling Techniques

Any dynamic system can be modeled by using at least one of the following four techniques:
(a) governing equations by the first laws of nature (input–output differential-equation
formulation), (b) state-space representations, (c) transfer function formulations, and (d)
block diagram representation. In this section the basic concepts of these modeling techniques
 6.2 System Modeling Techniques 399

are discussed. The derivation of the governing equations for several types of physical systems
has been covered in Chapters 3 to 5. The remaining three techniques are presented in Sections
6.3 to 6.5.
The governing equations constitute an essential component of a dynamic system model.
Typically, they are represented by differential or differential-algebraic equations. The deriv-
ation of the governing equations uses the fundamental laws of physics, the choice of which
depends on the nature of the dynamic system. For example, for mechanical systems they are
Newton’s laws, while for electric circuits they are Kirchhoff’s current and voltage laws.
Detailed discussions on the derivation of the governing equations of mechanical and elec-
trical systems are presented in Chapters 3 and 4, respectively.
The input–output differential-equation(s) system model may contain a single equation or a
set of coupled equations. If such a model is sufficiently simple and can be easily implemented
in software, no other representations are required for system-level analysis. Alas, for the
majority of dynamic systems, and, in particular, for compound systems that consist of
multiple subsystems of different nature, this is not the case. Hence, this sort of dynamic
system needs to be modeled as a block diagram, a transfer function, or a state-space represen-
tation to simplify and streamline the system-level analysis.
To demonstrate the process of system-level modeling and analysis consider the simplest
compound dynamic system: a controlled mechanical system shown in Example 6.2.1.

Example 6.2.1 The mechanical subsystem shown in Figure 6.2.1 (a) consists of a block of mass m,
supported by a spring of stiffness k and a linear damper of the damping coefficient c. The block is
subjected to an external force FðtÞ. The motion of the mass is controlled by a differential controller
(differentiator, where D ¼ d=dt). The sensor measures the displacement xðtÞ of the mass m. This
displacement is processed by a controller and a control force Fc ¼ μx_ is then applied to the mass.

x(t)
k
F(t)
m
Fs
c Sensor F(t)
. Fd m
μx
x Fc
μD

Differentiator (controller)
(a) (b)

Figure 6.2.1 Controlled dynamic system: (a) description and (b) free-body diagram
400 6 Combined Systems and System Modeling Techniques

A free-body diagram for the mass m is shown in Figure 6.2.1(b). The model of the mechanical
subsystem is then

x ¼ FðtÞ  ðFs þ Fd þ Fc Þ
m€ (1)

where Fs ; Fd ; and Fc represent the spring, damper, and controller forces respectively.
Recalling the expressions for spring and damper forces – Fs ¼ kx and Fd ¼ cx_ (see Chapter 3) –
substitute them into Eq. (1):

m€ _  Fc
x ¼ FðtÞ  ðkx þ cxÞ (2)

The controller model is straightforward:

Fc ¼ μx_ (3)

Thus, the system model is described by two equations: Eq. (2) and Eq. (3):

_  Fc
x ¼ FðtÞ  ðkx þ cxÞ
m€
(4)
Fc ¼ μx_

This model can be easily reduced into a single governing equation:

x þ ðc þ μÞx_ þ kx ¼ FðtÞ
m€ (5)

The system model as described by either the set of Eqs. (4) or Eq. (5), is simple, and does not
require different representations for system-level analysis. Nonetheless, recalling the fundamen-
tals of state-space representations, transfer function formulations, and block diagram construc-
tions presented in Chapter 2 (Sections 2.5.4 and 2.7), these three types of system model are derived
as follows.
State-Space Representation
x1 ¼ xðtÞ
Select the state variables as . Since only the displacement of the mass m is monitored,
_
x2 ¼ xðtÞ
the output is yðtÞ ¼ xðtÞ ¼ x1 . Then, the state equations are derived as
( (
x_ 1 ¼ x2 x_ 1 ¼ x2
) k cþμ 1
mx_ 2 þ ðc þ μÞx2 þ kx1 ¼ FðtÞ x_ 2 ¼  x1  x2 þ FðtÞ
m m m

and the full state representation is

(
x_ 1 ¼ x2
k cþμ 1
x_ 2 ¼  x1  x2 þ FðtÞ
m m m
y ¼ x1
_ ¼ AxðtÞ þ BuðtÞ
xðtÞ
Cast in matrix form, it becomes , where
yðtÞ ¼ CxðtÞ þ DuðtÞ
 6.2 System Modeling Techniques 401

2 3 !
 0 1 0
x
xðtÞ ¼ 1 ; uðtÞ ¼ FðtÞ; A¼4 k c þ μ 5; B¼ 1 ; C ¼ ð1 0 Þ; D ¼ 0
x2  
m m m

Transfer Function Formulation

To derive the transfer function of this system, the Laplace transform of Eq. (5) is taken, assuming
zero initial conditions:

ms2 X ðsÞ þ ðc þ μÞsX ðsÞ þ kX ðsÞ ¼ FðsÞ (6)

From Eq. (6) the system transfer function is derived as follows:

X ðsÞ 1
GðsÞ ¼ ¼
FðsÞ ms2 þ ðc þ μÞs þ k

Block Diagram
The simplest block diagram is obtained from the system transfer function, and is as shown in
Figure 6.2.2.

F(s) 1 X(s)
ms 2 + (c + μ )s + k

Figure 6.2.2

6.2.1 State-Space Representations

As discussed in Section 2.5.4, state equations are a set of coupled first-order ordinary
differential equations of the form described by Eq. (2.5.9). In mathematical literature this
representation is often called the Cauchy form. As mentioned earlier, this form is well suited
for implementation in software to investigate system behaviors in simulation.
The state-space representation of a system consists of two distinct parts as shown in Figure
6.2.3: state equations and output equations.
A state equation is a set of coupled first-order ordinary differential equations, where the
left-hand side only contains the first derivative of a state variable, and the right-hand side
does not contain any derivative term of a state variable.
Nonlinear state equations are described by Eq. (2.5.9), while linear state equations are
described by Eq. (2.5.10). Output equations are in the form of:
8
>
> y1 ¼ g1 ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; um Þ
>
<y2 ¼ g2 ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; um Þ
.. (6.2.1)
>
> .
>
:
yk ¼ gk ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; um Þ
402 6 Combined Systems and System Modeling Techniques

x1
u1 y1
x2
u2 State Output y2
equations equations
um yk
xn

Inputs State variables Outputs

Figure 6.2.3 State-space representation of a dynamic system

where the unknown functions yi ¼ yi ðtÞ; i ¼ 1; 2; . . . ; k are the specified outputs;

xi ¼ xi ðtÞ; i ¼ 1; 2; . . . ; n are state variables; ui ¼ ui ðtÞ; i ¼ 1; 2; . . . ; m are given inputs;
and g1 ; g2 ; . . . ; gk are linear or nonlinear functions of the state variables and inputs.
Hence, the state-space representation of a dynamic system that is described by n state
variables, and has m inputs and k outputs is as follows:
8
>
> x_ 1 ¼ f1 ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; um Þ
>_
< x 2 ¼ f2 ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; um Þ
State equations .. (6.2.2)
>
> .
>
:
x_ n ¼ fn ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; um Þ
8
>
> y1 ¼ g1 ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; um Þ
>
<y2 ¼ g2 ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; um Þ
Output equations ..
>
> .
>
:
yk ¼ gk ðx1 ; x2 ; . . . ; xn ; u1 ; u2 ; . . . ; um Þ

In this formulation functions f1 ; f2 ; . . . ; fn and g1 ; g2 ; . . . ; gk are of state variables

(xi ¼ xi ðtÞ; i ¼ 1; 2; . . . ; n) and inputs (ui ¼ ui ðtÞ; i ¼ 1; 2; . . . ; m); they may be linear or
nonlinear.
The scope of this book covers only these dynamic systems that have a linear state-space
representation. It is of the form:
8
>x_ 1 ¼ a11 x1 þ a12 x2 þ . . . þ a1n xn þ b11 u1 þ b12 u2 þ . . . þ b1m um
>
<x_ 2 ¼ a21 x1 þ a22 x2 þ . . . þ a2n xn þ b21 u1 þ b22 u2 þ . . . þ b2m um
State equations .. (6.2.3)
>
> .
:
x_ n ¼ an1 x1 þ an2 x2 þ . . . þ ann xn þ bn1 u1 þ bn2 u2 þ . . . þ bnm um
8
>
> y1 ¼ c11 x1 þ c12 x2 þ . . . þ c1n xn þ d11 u1 þ d12 u2 þ . . . þ d1m um
>
<y2 ¼ c21 x1 þ c22 x2 þ . . . þ c2n xn þ d21 u1 þ d22 u2 þ . . . þ d2m um
Output equations ..
>
> .
>
:
yk ¼ ck1 x1 þ ck2 x2 þ . . . þ ckn xn þ dk1 u1 þ dk2 u2 þ . . . þ dkm um

where aji , bji , cji , and dji are either known functions of time or constants.
 6.2 System Modeling Techniques 403

The linear state-space representation of a dynamic system (Eq. (6.2.3)) is often cast in a
matrix form:

_ ¼ AxðtÞ þ BuðtÞ
xðtÞ
(6.2.4)
yðtÞ ¼ CxðtÞ þ DuðtÞ

where xðtÞ and uðtÞ are the state and the input column vectors respectively, defined as
0 1 0 1
x1 ðtÞ u1 ðtÞ
Bx2 ðtÞ C B u2 ðtÞ C
B C B C
xðtÞn1 ¼B . C; uðtÞm1 ¼B . C
@ .. A @ .. A
xn ðtÞ um ðtÞ

while matrices A, B, C, and D are as follows:

2 3 2 3
a11 a12 . . . a1n b11 b12 . . . b1m
6a21 a22 . . . a2n 7 6b21 b22 . . . b2m 7
6 7 6 7
Ann ¼ 6 .. .. .. 7; Bnm ¼ 6 .. .. .. 7
4 . . ⋱ . 5 4 . . ⋱ . 5
an1 an2 . . . ann bn1 bn2 . . . bnm
2 3 2 3
c11 c12 . . . c1n d11 d12 . . . d1m
6c21 c22 . . . c2n 7 6d21 d22 . . . d2m 7
6 7 6 7
Ckn ¼ 6 .. .. .. 7; Dkm ¼ 6 .. .. .. 7
4 . . ⋱ . 5 4 . . ⋱ . 5
ck1 ck2 . . . ckn dk1 dk2 . . . dkm

and column vector yðtÞk1 is the output vector.

A square matrix Ann is called a system matrix, rectangular matrices Bnm and Ckn are the
input matrix and the state output matrix, respectively, and the matrix Dkm is called either the
control output matrix or the direct transmission matrix.
Obviously, a state-space representation of a system is not unique since it depends on the
choice of state variables and outputs.
The general rule of the assignment of state variables to system parameters was discussed in
Chapter 2 (see Section 2.5.4). It states that n state variables must be defined to convert an n-th
order differential equation into a set of state equations. For an equation

dn xðtÞ dn1 xðtÞ dxðtÞ

an ðtÞ n
þ an1 ðtÞ n1
þ . . . þ a1 ðtÞ þ a0 ðtÞxðtÞ ¼ uðtÞ (6.2.5)
dt dt dt

where ak ðtÞ are known functions, state variables are selected as follows:

dxðtÞ d2 xðtÞ dn1 xðtÞ

x1 ðtÞ ¼ xðtÞ; x2 ðtÞ ¼ ; x3 ðtÞ ¼ ; . . . ; x n ðtÞ ¼ (6.2.6)
dt dt2 dtn1
404 6 Combined Systems and System Modeling Techniques

This selection yields a set of ðn  1Þ state equations:

8
>
> x_ 1 ðtÞ ¼ x2 ðtÞ
>
>
>
> x_ 2 ðtÞ ¼ x3 ðtÞ
>
>
<x_ 3 ðtÞ ¼ x4 ðtÞ
.. (6.2.7)
>
> .
>
>
>
> x_ n2 ðtÞ ¼ xn1 ðtÞ
>
>
:
x_ n1 ðtÞ ¼ xn ðtÞ

Rewriting the original differential equation (Eq. (6.2.5)) in terms of the defined state variables:

an ðtÞx_ n ðtÞ þ an1 ðtÞxn ðtÞ þ an2 ðtÞxn1 ðtÞ þ . . . þ

(6.2.8)
þ a2 ðtÞx3 ðtÞ þ a1 ðtÞx2 ðtÞ þ a0 ðtÞx1 ðtÞ ¼ uðtÞ

We obtain the n-th state equation:

a0 ðtÞ a1 ðtÞ a2 ðtÞ

x_ n ðtÞ ¼  x1 ðtÞ  x2 ðtÞ  x3 ðtÞ  . . . 
an ðtÞ an ðtÞ an ðtÞ
(6.2.9)
an2 ðtÞ an1 ðtÞ 1
 xn1 ðtÞ  xn ðtÞ þ uðtÞ
an ðtÞ an ðtÞ an ðtÞ

Thus, state equations of such dynamic system are represented by Eqs. (6.2.7) and (6.2.9),
while the formulation of the output equations depends on the choice of outputs. For
example, if all state variables need to be monitored, the output vector yðtÞ is the same as
the state vector xðtÞ. Then, the familiar state-space representation in matrix form:

_ ¼ AxðtÞ þ BuðtÞ
xðtÞ
yðtÞ ¼ CxðtÞ þ DuðtÞ

becomes
2 3 0 1
0 1 0 ... 0 0
6 ... 7 B C 0
6
0 0 1 0
7 B C ..
6 .. .. .. .. 7 B C
. . . ⋱ . B C .
AðtÞ ¼ 6
6
7; B ¼ B
7 C
6 0 0 0 ... 1 7 B 0 C
4 a0 ðtÞ 5 B C
a1 ðtÞ a2 ðtÞ an1 ðtÞ @ 1 A
   ... 
an ðtÞ an ðtÞ an ðtÞ an ðtÞ an ðtÞ
2 3 0 1
1 0 ... 0 x1 ðtÞ
60 1 0 07 Bx2 ðtÞ C
B C
CðtÞ ¼ 6
4 ... ... .. 7 ; DðtÞ ¼ 0; xðtÞ ¼ B . C ¼ yðtÞ (6.2.10)
⋱ .5 @ .. A
0 0 ... 1 xn ðtÞ

For a dynamic system described with a set of coupled differential equations, a similar rule
for assigning state variable applies. For any independent variable xj ðtÞ in the set of governing
 6.2 System Modeling Techniques 405

equations, a component state vector is needed. This component state vector is a column
vector with the number of rows equal to the highest order of derivative of the considered
independent variable xj ðtÞ. The system state vector is a union of all the component vectors.
The matrices A, B, C, and D are derived using the same procedure as described above.

Example 6.2.2
Derive a state-space representation for a system described by the following differential equation:

d3 xðtÞ d2 xðtÞ dxðtÞ

3
þ a1 2
þ a2 þ a3 xðtÞ ¼ uðtÞ
dt dt dt

where uðtÞ is the system input, the desired output is

dxðtÞ
yðtÞ ¼ b1 xðtÞ þ b2
dt

and ai and bj are known coefficients.

Solution
Since the differential equation is of third order, three state variables are needed. In accordance
with the procedure described above, the state variables are selected as follows:
8
<x1 ¼ xðtÞ
_
x ¼ xðtÞ
: 2
x3 ¼ x€ðtÞ

Then, the first two state equations arise from the definition of state variables:
x_ 1 ¼ x2 ðtÞ
x_ 2 ¼ x3 ðtÞ

and the last one is derived by rewriting the original governing equation in terms of the chosen state
variables:
x_ 3 þ a1 x3 þ a2 x2 þ a3 x1 ¼ uðtÞ
x_ 3 ¼ a3 x1  a2 x2  a1 x3 þ uðtÞ

The output equation is

y ¼ b1 x1 þ b2 x2

Thus, system’s state-space representation is

8
<x_ 1 ¼ x2 ðtÞ
x_ ¼ x3 ðtÞ
: 2
x_ 3 ¼ a3 x1  a2 x2  a1 x3 þ uðtÞ
y ¼ b1 x1 þ b2 x2
406 6 Combined Systems and System Modeling Techniques

or in matrix form:
_ ¼ AxðtÞ þ BuðtÞ
xðtÞ
yðtÞ ¼ CxðtÞ þ DuðtÞ
2 3 0 1
0 1 0 0
A3  3 ¼ 4 0 0 1 5; B3  1 ¼ @0 A C 1  3 ¼ ðb1 b2 0 Þ; D¼0
a3 a2 a1 1
0 1
x1
xðtÞ ¼ @x2 A; uðtÞ and yðtÞ are scalars
x3

Example 6.2.3
Derive the state-space representation for a system described by the following set of coupled
differential equations:
8
>
> d2 xðtÞ dxðtÞ dzðtÞ
<a1 þ a2 þ a3 xðtÞ  a4  a5 zðtÞ ¼ u1 ðtÞ
dt2 dt dt
>
>
:a6 dzðtÞ þ a7 zðtÞ  a8 dxðtÞ ¼ u2 ðtÞ
dt dt

where u1 ðtÞ and u2 ðtÞ are the system inputs, the desired outputs are
8
< dxðtÞ
y1 ðtÞ ¼ b1 xðtÞ þ b2
dt
:
y2 ðtÞ ¼ b3 zðtÞ þ b4 u2 ðtÞ

and ai and bj are known coefficients.

Solution
This dynamic system is described by two independent variables: xðtÞ and zðtÞ. The highest order of
derivative for xðtÞ is 2, hence two state variables are needed. Similarly, for zðtÞ, a single state
variable (first-order derivative) is needed. Then, the system state vector is of the size 3 × 1, and the
_
state variables are x1 ¼ xðtÞ; x2 ¼ xðtÞ; x3 ¼ zðtÞ
From the definition of these state variables, the first state equation is simply obtained:

x_ 1 ¼ x2 (1)

To derive the remaining two state equations, the governing equations are rewritten in terms of
state variables:

fa1 x_ 2 þ a2 x2 þ a3 x1  a4 x_ 3  a5 x3 ¼ u1 ðtÞ (2)

 6.2 System Modeling Techniques 407

fa6 x_ 3 þ a7 x3  a8 x2 ¼ u2 ðtÞ (3)

From Eq. (3), we find the expression for x_ 3

a8 a7 1
x_ 3 ¼ x2  x3 þ u2 (4)
a6 a6 a6

Plugging this result (Eq. (4)) into Eq. (2) and performing the necessary algebraic operations, the
expression for x_ 3 is obtained:
 
a3 a4 a8 a2 a4 a7 a5 1 1
x_ 2 ¼  x1 þ  x2 þ  x3 þ u1 þ u2 (5)
a1 a1 a6 a1 a1 a6 a1 a1 a1 a6

The output equations in terms of state variables are

y1 ðtÞ ¼ b1 x1 þ b2 x2
(6)
y2 ðtÞ ¼ b3 x3 þ b4 u2 ðtÞ

Then, the system state-space representation consists of Eqs. (1), (4), (5), and (6) as follows:
8
>x_ 1 ¼ x2  
>
>
<x_ ¼  a3 x þ a4 a8  a2 x þ a4 a7  a5 x þ 1 u þ 1 u
2 1 2 3 1 2
State equations a1 a1 a6 a1 a1 a6 a1 a1 a1 a6
>
>
>
:x_ 3 ¼ a8 x2  a7 x3 þ 1 u2
a6 a6 a6
y1 ðtÞ ¼ b1 x1 þ b2 x2
Output equations
y2 ðtÞ ¼ b3 x3 þ b4 u2 ðtÞ

If desired, the derived state-space representation can be cast in the matrix form:
_ ¼ AxðtÞ þ BuðtÞ
xðtÞ
, where the matrices are as follows:
yðtÞ ¼ CxðtÞ þ DuðtÞ
2 3 0 1
0 1 0 0 0
6 a3 a4 a8  a2 a6 a4 a7  a5 a6 7 B1 1 C
6 7 B C
A33 ¼ 6 a a a a a 7; B32 ¼ Ba1 a1 a6 C
B
C;
6 1 1 6 1 6 7 B C
4 a8 a7 5 @ 1 A
0  0
a6 a6 a6
 
b1 b2 0 0 0
C 23 ¼ ; D22 ¼
0 0 b3 0 b4
0 1
x1  
@ A u1 y
xðtÞ ¼ x2 ; uðtÞ ¼ ; yð t Þ ¼ 1
u2 y2
x3
408 6 Combined Systems and System Modeling Techniques

While the above procedure of selecting state variables applies to any linear dynamic system,
it is useful to think about the physical meaning of state variables as parameters that describe
the system’s energy. For example, for mechanical systems discussed in Chapter 3, such state
variables are displacement (linear or angular) and velocity (linear or angular). The displace-
ment is used to describe the potential energy of the system, arising due to linear or torsional
1 1
springs ðV ¼ kx2 or V ¼ kT θ2 Þ. The velocity describes the kinetic energy of the system,
2 2
1 1
arising due to the motion of inertia elements ðT ¼ mx_ 2 or T ¼ IG θ_ Þ. For electrical elem-
2
2 2
ents, discussed in Chapter 4, the state variables are assigned only for the elements that store
energy. These state variables are current and voltage, which are both used to describe the
1 1
electrical energy ðE ¼ Cv 2 or E ¼ Li2 Þ.
2 2
For compound dynamic systems that are described by a set of differential equations, state
variables are assigned per subsystem, and the resulting state vector will be a union of the
subsystems’ state vectors. The procedure of deriving a state-space representation of a
compound system is similar to that illustrated by Example 6.2.3. The formulation of such
state-space representations will be discussed in more detail later in this chapter.

6.2.2 Transfer Function Formulations

As introduced in Chapter 2 (see Section 2.7.1), the transfer function of a time-invariant linear
system is defined as the ratio of the system output to system input in the Laplace domain,
assuming zero initial conditions (Eq. (2.7.1)).
Consider a dynamic system described by a governing equation

d n xð t Þ dn1 xðtÞ dxðtÞ

an n
þ an1 n1
þ . . . þ a1 þ a0 xðtÞ ¼
dt dt dt
dm uðtÞ dm1 uðtÞ duðtÞ
¼ bm þ bm1 þ . . . þ b1 þ b0 uðtÞ (6.2.11)
dtm dtm1 dt

where ai and bj are known coefficients, uðtÞ is the system input, and xðtÞ is the system output.
Assuming zero initial conditions, take the Laplace transform of Eq. (6.2.11):
 
an sn þ an1 sn1 þ . . . þ a1 s þ a0 X ðsÞ ¼
  (6.2.12)
¼ bm sm þ bm1 sm1 þ . . . þ b1 s þ b0 UðsÞ

Then the sought system transfer function is

X ðsÞ bm sm þ bm1 sm1 þ . . . þ b1 s þ b0
GðsÞ ¼ ¼ (6.2.13)
U ðsÞ an sn þ an1 sn1 þ . . . þ a1 s þ a0

Recalling the discussion in Section 2.4.2, the poles of this transfer function are defined as
the solutions of the equation
 6.2 System Modeling Techniques 409

an sn þ an1 sn1 þ . . . þ a1 s þ a0 ¼ 0 (6.2.14)

and its zeros are the solutions of the equation

bm sm þ bm1 sm1 þ . . . þ b1 s þ b0 ¼ 0 (6.2.15)

Obviously, the derived transfer function has n poles and m zeros.

Knowledge of the poles and zeros of the system transfer function gives a good insight into
system behavior; these parameters are widely used in control design. The order of a transfer
function is the number of its poles; hence, the transfer function described by Eq. (6.2.13) is of
n-th order.
A transfer function can be expressed in a zero-pole-gain form such as

X ðsÞ ðs  z1 Þðs  z2 Þ . . . ðs  pn Þ
GðsÞ ¼ ¼k (6.2.16)
U ðsÞ ðs  p1 Þðs  p2 Þ . . . ðs  zm Þ

bm
where pi and zj are the transfer function poles and zeros respectively, and k ¼is the gain. A
an
pole-zero-gain formulation is also extensively used in system analysis and control design.
Since the transfer function form depends on the choice of input and output for the
dynamic system, it is not unique.
For any linear time-invariant dynamic system, its transfer function can be obtained from
the system state-space representation cast in matrix form:

GðsÞ ¼ CðsI  AÞ1 B þ D (6.2.17)

where the matrices A, B, C, and D are defined as shown by Eq. (6.2.4), and I is the identity
matrix of the same size as the state matrix A.

Example 6.2.4
Derive transfer functions, find their poles and zeros, and cast the obtained transfer functions into
pole-zero-gain form for the dynamic systems described by the following governing equations:

(a) x_ ðtÞ þ axðtÞ ¼ uðtÞ

(b) x€ðtÞ þ a1 x_ ðtÞ þ a0 xðtÞ ¼ uðtÞ
Ðt
(c) x€ðtÞ þ a1 xðtÞ þ a0 xðτÞdτ ¼ b1 uðtÞ
_ þ b0 uðtÞ
0

Solution
For the system input uðtÞ and output xðtÞ, the transfer functions are obtained by taking the
Laplace transform of the governing equations, followed by the algebraic operations as follows:

(a) L½x_ ðtÞ þ axðtÞ ¼ L½uðtÞ

410 6 Combined Systems and System Modeling Techniques

ðs þ aÞX ðsÞ ¼ U ðsÞ

X ðsÞ 1
GðsÞ ¼ ¼
U ðsÞ s þ a

This is the first-order transfer function (n ¼ 1; m ¼ 0), with a single pole p1 ¼ a and no zeros.
It is already in the pole-zero-gain form (an ¼ 1; bm ¼ 1).

(b) L½€x ðtÞ þ a1 x_ ðtÞ þ a0 xðtÞ ¼ L½uðtÞ

2 
s þ a1 s þ a0 X ðsÞ ¼ UðsÞ
X ðsÞ 1
GðsÞ ¼ ¼
U ðsÞ s2 þ a1 s þ a0

This is the second-order transfer function (n ¼ 2; m ¼ 0), with two poles p1 and p2 that are the
roots of the equation s2 þ a1 s þ a0 ¼ 0 and no zeros. The pole-zero-gain form of transfer function
is
1
GðsÞ ¼
ðs  p1 Þðs  p2 Þ
" #
Ðt h i
(c) L x€ðtÞ þ a1 xðtÞ þ a0 xðτÞdτ ¼ L b1 u_ ðtÞ þ b0 uðtÞ
0
 a0 
s2 þ a1 þ X ðsÞ ¼ ðb1 s þ b0 ÞUðsÞ
s

X ðsÞ b1 s þ b0 b1 s2 þ b0 s
GðsÞ ¼ ¼ ¼
U ðsÞ s2 þ a1 þ a0 s3 þ a1 s þ a0
s

This is the third-order transfer function (n ¼ 3; m ¼ 2), with three poles p1 , p2 , and p3 that are
the roots of the equation s3 þ a1 s þ a0 ¼ 0 and two zeros z1 ¼ b0 =b1 and z2 ¼ 0.
Note that any transfer function must be a ratio of two polynomials. Hence, the form

b1 s þ b0
GðsÞ ¼
s2 þ a1 þ a0 =s

is not a proper transfer function since its denominator is not a polynomial. Multiplying the
numerator and denominator by s yields the sought transfer function, but with an additional
zero (z2 ¼ 0) and of order greater than the order of the highest derivative of xðtÞ.
The pole-zero-gain form of the derived transfer function is

sðs þ b0 =b1 Þ
GðsÞ ¼ b1
ðs  p1 Þðs  p2 Þðs  p3 Þ
 6.2 System Modeling Techniques 411

Dynamic systems with multiple inputs and multiple outputs (MIMO systems) have a
transfer matrix, the elements of which are transfer functions that reflect every input–output
relation in the system. The derivation of these transfer functions is done using the same
procedure as described above. Every output Yi ðsÞ is expected to yield an expression of the
form

Yi ðsÞ ¼ G1i ðsÞU1 ðsÞ þ G2i ðsÞU2 ðsÞ þ . . . þ Gðn1Þi ðsÞUðn1Þ ðsÞ þ Gni ðsÞUn ðsÞ

where Gji ðsÞ denotes the transfer function from the j-th input Uj ðsÞ to the i-th output Yi ðsÞ. An
individual transfer function Gji ðsÞ is derived by setting all the inputs other than Uj ðsÞ to zero,
thus obtaining an expression Yi ðsÞ ¼ Gji ðsÞUj ðsÞ.
Upon deriving all the transfer functions, the transfer matrix for a system with n inputs and
m outputs is constructed. It is of the form
2 3
G11 ðsÞ G21 ðsÞ . . . Gn1 ðsÞ
6 G12 ðsÞ G22 ðsÞ . . . Gn2 ðsÞ 7
6 7
Gmn ðsÞ ¼ 6 . .. .. 7
4 .. . ⋱ . 5
G1m ðsÞ G2m ðsÞ ... Gnm ðsÞ

Verifying that multiplication of this matrix by the n × 1 input vector produces the correct
expressions for every output Yi ðsÞ is straightforward.
Even for the low-order MIMO systems, the derivation of the transfer matrix requires a
number of algebraic operations and may become quite tedious.

Example 6.2.5
Derive a transfer matrix for the two-input–two-output system described by the following set of
governing equations:

_ þ a1 zðtÞ ¼ u1 ðtÞ
xðtÞ
z€ðtÞ þ a2 z_ ðtÞ þ a3 xðtÞ ¼ u2 ðtÞ

System inputs are u1 ðtÞ and u2 ðtÞ, and the outputs are the independent variables xðtÞ and zðtÞ.
Solution
Since this dynamic system has two inputs and two outputs, four transfer functions are expected, as
shown in Figure 6.2.4.
412 6 Combined Systems and System Modeling Techniques

U1(s) X(s)
G1x
G1z G 2x
U2(s)
Z(s)
G 2z

Dynamic system

Figure 6.2.4 Transfer function representation of a two-input–two-output dynamic system

Let the system input vector be


U1 ðsÞ
UðsÞ ¼
U2 ðsÞ

and output vector be


X ðsÞ
YðsÞ ¼
ZðsÞ

Recalling that YðsÞ ¼ GðsÞUðsÞ, the system transfer matrix has the following form:

G1X G2X
GðsÞ ¼
G1Z G2Z

where G1X , G2X , G1Z , and G2Z are the individual transfer functions relating every input to every
output as shown in Figure 6.2.4.
As discussed earlier, the sought transfer functions are derived taking the Laplace transform of
the governing equations assuming zero initial conditions, and then performing the necessary
algebraic transformations.
Taking the Laplace transform of the given governing equations, we obtain

fsX ðsÞ þ a1 ZðsÞ ¼ U1 ðsÞ (1)

n 
s2 þ a2 s ZðsÞ þ a3 X ðsÞ ¼ U2 ðsÞ (2)

To derive the transfer functions for the X ðsÞ output, i.e. G1X and G2X , we need to have a single
equation in terms of this output and the inputs. This is done by deriving an expression for ZðsÞ
from Eq. (1):

s 1
ZðsÞ ¼  X ðsÞ þ U1 ðsÞ (3)
a1 a1

and substituting Eq. (3) into Eq. (2), obtaining


s 1
ðs þ a2 sÞ  X ðsÞ þ U1 ðsÞ þ a3 X ðsÞ ¼ U2 ðsÞ
2
(4)
a1 a1
 6.2 System Modeling Techniques 413

Simplifying Eq. (4) yields

s2 þ a2 s a1
X ðsÞ ¼ U1 ðsÞ  3 U2 ðsÞ (5)
s þ a2 s  a1 a3
3 2 s þ a2 s2  a1 a3

Hence, the transfer functions G1X and G2X are found by setting the inputs in Eq. (5) to zero in
turn:

X ðsÞ s2 þ a2 s
G1X ¼ ¼ 3
U1 ðsÞ s þ a2 s2  a1 a3
(6)
X ðsÞ a1
G2X ¼ ¼ 3
U2 ðsÞ s þ a2 s2  a1 a3

Similarly, we derive the transfer functions G1Z and G2Z .

From Eq. (2):

s2 þ a2 s 1
X ðsÞ ¼  ZðsÞ þ U2 ðsÞ (7)
a3 a3

Substituting Eq. (7) into Eq. (1), we derive an expression for ZðsÞ:
 2
s þ a2 s 1
s  ZðsÞ þ U2 ðsÞ þ a1 ZðsÞ ¼ U1 ðsÞ
a3 a3
(8)
a3 s
ZðsÞ ¼  3 U1 ðsÞ þ 3 U2 ðsÞ
s þ a2 s2  a1 a3 s þ a2 s2  a1 a3

Eq. (8) yields the transfer functions G1Z and G2Z :

ZðsÞ a3
G1Z ¼ ¼ 3
U1 ðsÞ s þ a2 s2  a1 a3
(9)
ZðsÞ s
G2Z ¼ ¼
U2 ðsÞ s3 þ a2 s2  a1 a3

Having derived the component functions, the system transfer matrix is easily constructed:
2 3
s2 þ a2 s a1
6 s3 þ a2 s2  a1 a3 
G1X G2X s3 þ a2 s2  a1 a3 7
GðsÞ ¼ ¼6
4
7
5
G1Z G2Z a3 s
 3
s þ a2 s2  a1 a3 s þ a2 s  a1 a3
3 2
414 6 Combined Systems and System Modeling Techniques

6.2.3 Block Diagrams

Block diagrams, as introduced in Chapter 2 (see Section 2.7.2) provide a powerful method
for describing the input–output relations within a dynamic system. It is particularly useful
for MIMO systems as well as compound systems, even though at the system level a
compound system may have a single input and a single output. A block diagram representa-
tion is a very general cause-and-effect approach to dynamic system modeling and can be
applied to many types of systems – linear and nonlinear, time-varying and time-invariant.
Additionally, constructing a block diagram for the given dynamic system helps in modeling
this system in software such as Simulink by MathWorks, or System Modeler by Wolfram
Research.
A block diagram has three major components: the block, summation point, and branch or
takeoff point, as illustrated in Figure 6.2.5.
In Figure 6.2.5 the arrows represent variables (signals) and the direction of the input–
output (cause-and-effect) relation. A block, presented in Figure 6.2.5(a), embodies the
transfer function between the depicted input X ðsÞ and output YðsÞ. Generally, for a dynamic
system described by a set of differential equations, the block represents a single equation.
Figure 6.2.5(b) shows the summation point. Depending on the signs at the variables entering
the summation point (plus or minus at an arrowhead), either addition or subtraction
operations may be performed. A summation point may have more than two incoming
signals. Figure 6.2.5(c) shows the branch point, from which the signal goes concurrently to
the other elements of the block diagram. Note that the branch point does not change the
value of a variable that entered it, but only conveys it to the other parts of this block diagram.
It may be desirable to describe a linear time-invariant system that has a state-space
representation

_ ¼ AxðtÞ þ BuðtÞ
xðtÞ
yðtÞ ¼ CxðtÞ þ DuðtÞ

(see Eq. (6.2.4)) by a block diagram. There, the blocks describe input–output relations using
matrices A, B, C, D, and an integrator rather than a transfer function (see Figure 6.2.6).

X 1(s)
X(s)
X(s) Y(s) Y(s) + X 1 (s)+X 2 (s) X(s)
X(s) X(s)
+
X(s)
X 2(s)

(a) (b) (c)

Figure 6.2.5 Basic components of a block diagram: (a) block, (b) summation point, (c) branch point
 6.2 System Modeling Techniques 415

Figure 6.2.6 Block diagram of a linear time-invariant dynamic system

A block diagram represents a powerful visual depiction of a dynamic system. It clearly

shows how parameters evolve throughout the system, and how system components interact
with each other. Nonetheless, for all but the simplest dynamic systems, the block diagram
often becomes a convoluted structure, with a large number of blocks, summation, and
branch points, and a tangled web of arrows. A complicated block diagram can be simplified
by applying the following rules of equivalent transformation:

(a) a combination of two blocks can be substituted with a single block, the transfer function
of which is computed as shown in Table 6.2.1, where G1 and G2 denote the transfer
functions of connected blocks, while UðsÞ and YðsÞ are the input and output of the block
combination, respectively
(b) a branch point can be moved across a block, but never across a summation point;
movement of a branch point across any block is accompanied by the addition of an
appropriate block element to prevent the loss of the signal

The application of these rules to simplify a block diagram is illustrated in Examples 6.2.7
through 6.2.10.
It needs to be noted that series and parallel connections may include more than two blocks.
The equivalent transformation of n blocks in series is still a single block with the transfer
function

Geq ¼ ∏ Gi (6.1.18)
n

where Gi denotes a transfer function of an individual block component.

Similarly, for n blocks in parallel the equivalent block transfer function is
X
Geq ¼ Gi (6.1.19)
n

The derivation of the equivalent transfer functions presented in Table 6.2.1, and expressed
with Eqs. (6.1.18) and (6.1.19), is straightforward.
416 6 Combined Systems and System Modeling Techniques

Table 6.2.1 Equivalent transformation of block combinations

Configuration Equivalent Transformation

1. Series (cascade) connection

U(s) Y(s) U(s) Y(s)

G1 G2 G1G2

2. Parallel connection

G1
U(s) + Y(s) U(s) Y(s)
G1+G2
+
G2

3. Feedback connection (negative feedback)

U(s) + Y(s)
G1 U(s) G1 Y(s)
–
1+G1G2
G2

4. Feedback connection (positive feedback)

U(s) + Y(s)
G1 U(s) G1 Y(s)
+ 1–G1G2
G2

5. Feedback connection (negative unity

feedback: transfer function of a simple
connector line equals 1)
U(s) G1 Y(s)
U(s) + Y(s)
G1 1+G1
–

Example 6.2.6
Derive the equivalent transfer function of a negative feedback configuration shown in Figure
6.2.7.

Solution
The first step in the derivation includes denoting signals that travel along each shown arrow. Let
the outgoing signal of the summation point (in control theory it is often called the error signal) be
denoted as EðsÞ ¼ X ðsÞ  G2 YðsÞ. Then, the expression for the output signal, shown in Figure
6.2.7, can be rewritten as follows:
 6.2 System Modeling Techniques 417

X(s ) + E(s) G1 E(s ) = Y(s )

G1
–

G2Y(s ) Y(s)
G2

Figure 6.2.7 System for Example 6.2.6

YðsÞ ¼ G1 EðsÞ ¼ G1 ðX ðsÞ  G2 YðsÞÞ

Rearranging, we obtain
YðsÞð1 þ G1 G2 Þ ¼ G1 X ðsÞ

Thus, the sought transfer function is

YðsÞ G1
GðsÞ ¼ ¼
X ðsÞ 1 þ G1 G2

Example 6.2.7
Consider a dynamic system, modeled with the block diagram presented in Figure 6.2.8.

Solution
While this block diagram is not exceedingly complex, deriving the system transfer function from it
requires an effort. Hence, constructing an equivalent block diagram is necessary. Since none of the
block combinations shown in Table 6.2.1 are discernible, the first simplification step encompasses
moving the branch point A across the block G2 . This will yield the series connection of the blocks
G1 and G2 , and the parallel connection of the block G4 with the elements located on the line
connecting branch point A and the summation point. Remembering that the signal continuity
must be preserved with an equivalent transformation, the signals traveling between the blocks G2 ,
G3 , and G4 need to be studied prior to the equivalent transformation.
Let the signal entering the block G2 be denoted z. Then, the distribution of signals between the
blocks G2 , G3 , and G4 is as shown in Figure 6.2.9. The depicted input and output signals for every
block must be preserved in an equivalent transformation. This is achieved by adding a block with
the transfer function inverse to that of the block G2 , as illustrated in Figure 6.2.10.

U(s ) + A B Y(s )
G1 G2
–

G3 G4

+ +

Figure 6.2.8 System for Example 6.2.7

418 6 Combined Systems and System Modeling Techniques

As seen in Figure 6.2.10, the addition of the block with the transfer function 1=G2 preserves the
input signal of the block 1=G3 . Hence, it can be deduced that moving a branch point across a block
always entails the addition of a compensating block with the inverse transfer function right after
the moved branch point on the relocated line.
Moving the branch point A across the block G2 yields two recognizable series connections: one
of blocks G1 and G2 , and the other one of the block G3 and the newly added block 1=G2 . The
equivalent transformation of these combinations results in the block diagram depicted in Figure
6.2.11.
The equivalent transformation of the shown parallel connection of two blocks G4 and G3 =G2
yields the system model in standard negative feedback configuration, illustrated in Figure 6.2.12.
This block diagram is very suitable for implementation in software and subsequent simulation
and analysis. Nonetheless, if desired, it can be simplified even further, yielding the system transfer
function shown in Figure 6.2.13.
As demonstrated by this example, a dynamic system can be modeled by many equivalent block
diagrams. With the given input and output of the dynamic system, the shape of the sought block
diagram representation depends on which parameters and component relations need to be
studied.

A z zG2 B
z G2
z zG2

G3 G4

zG3 zG2G4

Figure 6.2.9 Distribution of signals

U(s ) + z zG2 A B Y(s )

G1 G2
– zG2 zG2
Series
1 G4
G2

G3
zG2G4
zG3
Series
+ +

Figure 6.2.10 First equivalent transformation

 6.2 System Modeling Techniques 419

U(s ) + A B Y(s)
G1G2
–
G3
G4
G2

+ +

Parallel

Figure 6.2.11 Second equivalent transformation

U(s ) + Y(s)
G1G2
–
G3
G2
+ G4

Negative feedback

Figure 6.2.12 Third equivalent transformation

Figure 6.2.13 System transfer function

Example 6.2.8
Consider a typical dynamic system, described by a second-order differential equation with the constant
coefficients x€ðtÞ þ axðtÞ
_ þ bxðtÞ ¼ f ðtÞ. Let the system input be f ðtÞ and the system output be xðtÞ.
_
Consider zero initial conditions: xð0Þ ¼ 0 and xð0Þ ¼ 0.

Solution
The system transfer function GðsÞ ¼ X ðsÞ=FðsÞ is easily derived by taking the Laplace transform
of the governing equation and performing algebraic transformations as follows:
420 6 Combined Systems and System Modeling Techniques

s2 X ðsÞ þ asX ðsÞ þ bX ðsÞ ¼ FðsÞ

X ðsÞðs2 þ as þ bÞ ¼ FðsÞ
X ðsÞ 1
¼ 2
FðsÞ s þ as þ b

If we are interested only in the input–output relation at the system level, the block diagram
representation is the simplest and most concise one. It utilizes the derived above system transfer
function and has only a single block, as shown in Figure 6.2.14.
There are several options of the block diagram representation better suitable for studying the
influence of the parameters x and x_ on the system dynamics. Rearranging the original governing
equation and taking the Laplace transform to construct these block diagrams, we obtain

x€ðtÞ ¼ axðtÞ
_  bxðtÞ þ f ðtÞ
s2 X ðsÞ ¼ asX ðsÞ  bX ðsÞ þ F ðsÞ
1  1 1  
X ðsÞ ¼ 2 asX ðsÞ  bX ðsÞ þ F ðsÞ ¼ asX ðsÞ  bX ðsÞ þ F ðsÞ
s s s

From these expressions the three-element summation point is easily discernible. Recalling that
in the Laplace domain x_ becomes sX ðsÞ, the block diagram shown in Figure 6.2.15 is constructed.
If a standard feedback configuration is desired, the inner feedback loop of blocks 1=s and a can
be simplified as specified in Table 6.2.1. The resulting block diagram is shown in Figure 6.2.16.
Thus, like the state-space representation of a dynamic system, this block diagram model is not
unique. The shape of the system’s block diagram depends on the system-level inputs and outputs
as well as on the elements in which the dynamic behavior needs to be studied.

Figure 6.2.14 Second-order system in Example 6.2.8

Figure 6.2.15 Expanded block diagram of a second-order system

 6.2 System Modeling Techniques 421

Figure 6.2.16 Second-order system in standard feedback configuration

A block diagram representation is particularly useful for MIMO dynamic systems. As

discussed in Section 6.2.2, these systems have a transfer matrix that contains the transfer
functions for every input–output relation in the system. While the component transfer
functions can be derived directly from the Laplace transform of the governing equations,
such derivations involve multiple algebraic operations and are often tedious. Obtaining these
transfer functions from a block diagram is much easier. The transfer function from input Ui
to output Yj is obtained by setting all the other inputs to zero and simplifying the block
diagram accordingly. The alternative approach includes processing the individual signals to
yield the expression for the desired output in terms of all inputs, and then setting all but one
inputs to zero.

Example 6.2.9
Consider a two-input–one-output dynamic system represented by the block diagram shown in
Figure 6.2.17. Derive the transfer functions

YðsÞ
GU ðsÞ ¼
UðsÞ

and

YðsÞ
GV ðsÞ ¼
VðsÞ

Let the outgoing signals of the summation points A and B be EðsÞ and WðsÞ, respectively.
According to the block diagram, these signals are

U(s) + A E(s ) G1E(s ) = Y(s )

G1
–
+ + V(s )
G2
G2W(s ) W(s) B

Figure 6.2.17 Two-input–one-output dynamic system for Example 6.2.9

422 6 Combined Systems and System Modeling Techniques

 
EðsÞ ¼ U ðsÞ  G2 W ðsÞ ¼ U ðsÞ  G2 V ðsÞ þ Y ðsÞ (2)

Substituting this expression (Eq. (2)) into the equation for YðsÞ shown in the block diagram
(Figure 6.2.17), we obtain
h  i
Y ðsÞ ¼ G1 EðsÞ ¼ G1 U ðsÞ  G2 V ðsÞ þ Y ðsÞ (3)

which, after rearranging to separate YðsÞ yields

G1 G1 G2
YðsÞ ¼ UðsÞ  VðsÞ (4)
1 þ G1 G2 1 þ G1 G2

From Eq. (4) the sought transfer functions are clearly seen:

G1
GU ðsÞ ¼
1 þ G1 G2

and
G1 G2
GV ðsÞ ¼ 
1 þ G1 G2

Setting the inputs to zero in turn and simplifying the block diagram accordingly yields this
solution with less work.
When VðsÞ ¼ 0, the block diagram becomes a standard negative feedback connection of two
blocks G1 and G2 (see Figure 6.2.18(a)). Thus, the equivalent transfer function is found as

G1
GU ðsÞ ¼
1 þ G1 G2

When UðsÞ ¼ 0, the block diagram becomes a positive unity feedback configuration, where the main
line contains blocks G1 and G2 connected in series. In the original block diagram, the signal originating
from the summation point B is transferred to the input of the block G1 with a negative sign. That sign
needs to be preserved in the equivalent transformation. Hence, as seen in Figure 6.2.18(b), the transfer
function of the block G2 assumes negative sign. The equivalent transfer function is then found as

G1 G2
GV ðsÞ ¼ 
1 þ G1 G2

U(s) + A E(s ) G1
Y(s ) V(s ) + B W(s) Y(s )
–G2 G1
– +

G2

(a) (b)

Figure 6.2.18 Derivation of transfer functions: (a) GU ðsÞ and (b) GV ðsÞ
 6.2 System Modeling Techniques 423

Example 6.2.10
Revisit the two-input two-output dynamic system described in Example 6.2.5, construct its block
diagram, and derive the system transfer matrix using the block diagram.

Solution
The system equations in the Laplace domain were derived as

fsX ðsÞ þ a1 ZðsÞ ¼ U1 ðsÞ (1)

n 
s2 þ a2 s ZðsÞ þ a3 X ðsÞ ¼ U2 ðsÞ (2)

A block diagram for a system described by a set of equations is constructed by parts, every part
corresponding to a single equation. The final step encompasses connecting the individual parts
together.
The block diagram for Eq. (1) is shown in Figure 6.2.19(a), and that for Eq. (2) is in Figure
6.2.19(b).
Connecting these partial diagrams, we obtain the system block diagram as illustrated in Figure
6.2.20.
Setting the inputs to zero in turn, we derive the component transfer functions. When U1 ðsÞ ¼ 0,
the block diagram is in negative feedback configuration, having on the main line a single transfer
1
function 2 , and three blocks in series on the feedback line, as shown in Figure 6.2.21(a).
s þ a2 s
Simplifying the diagram by substituting the series connection with a single block

U1(s ) + 1
X(s) U2(s ) + 1 Z(s )
s s 2 + a 2s
– –
a 1Z(s ) a 3X(s )
(a) (b)

Figure 6.2.19 Component block diagrams for the equations in Example 6.2.10: (a) Eq. (1), (b) Eq. (2)

U1(s ) + A X(s )
1
s
–
a1Z(s )

a3
a1

a3X(s )

Z(s ) 1 – + U2(s )

B s2 + a 2s

Figure 6.2.20 System block diagrams

424 6 Combined Systems and System Modeling Techniques

a1 a3
Gfb ¼
s3 þ a2 s2

and moving the branch point B toward the branch point A, we obtain the block diagram shown in
Figure 6.2.21(b).
Setting U2 ðsÞ ¼ 0 yields the block diagram shown in Figure 6.2.22(a), which is then similarly
processed to obtain the block diagram in Figure 6.2.22(b).
Using the equivalent transformations presented in Table 6.2.1, we obtain the sought transfer
functions:

X ðsÞ s2 þ a2 s
G1X ¼ ¼ 3
U1 ðsÞ s þ a2 s2  a1 a3
ZðsÞ a3
G1Z ¼ ¼ 3
U1 ðsÞ s þ a2 s2  a1 a3
X ðsÞ a1
G2X ¼ ¼ 3
U2 ðsÞ s þ a2 s2  a1 a3
ZðsÞ s
G2Z ¼ ¼
U2 ðsÞ s3 þ a2 s2  a1 a3

U2(s) + 1 B Z(s ) X(s )

s 2 + a2s
– a1
s
–a1
a3 U2(s ) + 1 Z(s )
s 2 + a2s
+
X(s )
1 a1a 3
s
A s

(a) (b)

Figure 6.2.21 Equivalent transformations when U1 ðsÞ ¼ 0 (see text for details)

Figure 6.2.22 Equivalent transformations when U2 ðsÞ ¼ 0 (see text for details)
 6.3 Fundamentals of Electromechanical Systems 425

6.3 Fundamentals of Electromechanical Systems

An electromechanical system, as its name suggests, consists of two coupled subsystems:
electrical and mechanical. The structure of such a system is illustrated in Figure 6.3.1.
An electrical subsystem is typically represented by an active or passive circuit, while a
mechanical subsystem may contain all three types of basic elements discussed in Chapter 3:
inertia, elastic, and damping components (masses, springs, and dampers).
As seen in Figure 6.3.1, energy conversion takes place in any electromechanical system.
Either the electrical subsystem generates a force to move the inertial elements of the
mechanical subsystem, or the motion of these masses generates current or voltage to feed
the electrical subsystem. The former case is exemplified by a DC motor, where an applied
voltage serves as an input and the parameters describing the rotation of the motor, such as
motor torque and angular velocity, are system outputs. The latter case is observed in
electroacoustic devices such as a microphone, where the motion of an elastic membrane
subject to sound waves generates a voltage for the electrical circuit.

6.3.1 Basic Concepts of Electromagnetism

Typically, coupling between mechanical and electrical subsystems is achieved via electro-
magnetic effects. Hence, the two fundamental principles of electromagnetism need to be
reviewed prior to an exploration of the approaches to modeling of the electromechanical
systems. The first principle deals with the force exerted by a magnetic field on the current-
carrying wire placed inside this field. The second one concerns the voltage induced in a
conductor moving in a magnetic field.
It was found experimentally that when a charged particle moves in a magnetic field,
characterized by the flux density ~B (also called magnetic field strength), this field exerts a
magnetic force ~ f on that charge. This force depends on the charge q of the particle, its
velocity ~
v, and flux density of the field as follows:

Generates force
i(t ), v(t ) to move masses F(t )

Electrical Mechanical
Energy conversion
subsystem subsystem

.
Motion of mass x(t ), x(t )
i(t ), v(t ) ..
generates x(t )
current or voltage

Figure 6.3.1 Typical electromechanical system

426 6 Combined Systems and System Modeling Techniques

~ v ~
f ¼ qð~ BÞ (6.3.1)

Since the electromagnetic force is a vector product of a particle’s velocity and field flux
density, it is perpendicular to the plane spanned by the vectors ~B and ~v, and its direction is
found in accordance with the right-hand rule (see Section 2.1.4). Note that the right-hand rule
applies to a positive charge; thus, for a moving negative charge, the velocity vector needs to
be reversed prior to using this rule.
According to the definition of a vector product, the magnitude of this magnetic force is

f ¼ qvB sin θ (6.3.2)

where θ is the angle between the vectors ~

B and ~v; 0° ≤ θ ≤ 180°.
Obviously, for the magnetic force to be non-zero, the charged particle must cross the
magnetic field lines while moving, i.e. vectors ~
B and ~v are not parallel (θ ≠ 0° and θ ≠ 180°).
Equation (6.3.1) is derived from the Lorentz force law:

~
F ¼ q~ v ~
E þ qð~ BÞ (6.3.3)

where ~F is the total force exerted by an electromagnetic field of electric strength ~

E and
~ ~
magnetic flux density B on a charged particle q; qE is the electric force.
The magnetic force exerted by a magnetic field on a current-carrying wire is derived from
Eq. (6.3.1) by considering the constrained motion of a charge element dq in a wire. The
velocity of this charge element is

d~l
~
v¼ (6.3.4)
dt

where d~l is an element of the length of the wire. The direction of this vector is the same as the
direction of the electrical current in the wire.
Then, the magnetic force acting on this charge element is
!
d~l dq
d~
f ¼ dqð~v ~ BÞ ¼ dq B ¼ ðd~l  ~
~ BÞ (6.3.5)
dt dt

Recalling the definition of electrical current (Eq. (4.1.1)), the above expression for the
magnetic force is rewritten as

d~
f ¼ iðd~l  ~
BÞ (6.3.6)

The entity i d~
l is often called the current element, which is a vector with the magnitude i
and the direction of the electrical current. Integrating the current element for the finite length
 6.3 Fundamentals of Electromechanical Systems 427

of the wire (0 ≤ l ≤ L) yields the vector L~i. Obviously, its direction is the direction of the
current in the wire, and its magnitude equals Li.
To obtain the magnetic force acting on the wire of length L, Eq. (6.3.6) is integrated for
0 ≤ l ≤ L and becomes

~
f ¼ L~i  ~
B (6.3.7)

The direction of the magnetic force is obtained using the right-hand rule as illustrated in
Figure 6.3.2, Since the conventional current represents motion of positively charged par-
ticles, the application of the right-hand rule is straightforward: when the index finger points
in the direction of current~i, and middle finger points in the direction of the magnetic field flux
density ~ B, the thumb will indicate the direction of the exerted magnetic force ~f.
The magnitude of the force, exerted by the magnetic field on the current-carrying wire, is

f ¼ LiB sin θ (6.3.8)

where θ is the angle between the vectors ~

B and ~i; 0° ≤ θ ≤ 180°.
Similarly to the case depicted in Figure 6.3.2, for the systems discussed further in this
chapter we assume that the flux density is uniform, the wire is straight, and that the flux
density and current vectors are perpendicular. Hence, the expression for the magnetic force
becomes

f ¼ LBi (6.3.9)

Equation (6.3.9) also applies to the circular wires in a radial magnetic field.
In the SI measurement system the magnetic flux density is measured in webers per square
meter (Wb/m2), while wire length is in meters (m) and current is in amperes (A). The unit of
magnetic flux is also called the tesla (T): T ¼ Wb=m2 ¼ N=A  m
A special case of interest represents a current-carrying rectangular wire loop in a uniform
magnetic field, as shown in Figure 6.3.3. The axis of the rectangular wire loop N1 N2 N3 N4 is

(a) (b)

Figure 6.3.2 Right-hand rule: (a) a current-carrying wire in the field of a permanent magnet with the
polarity shown; (b) using the right-hand rule to find the direction of the exerted magnetic force
428 6 Combined Systems and System Modeling Techniques

f2–3

i N2
B
N3 f1–2

B i

f3–4 N1

N4

– vin +

Figure 6.3.3 Current-carrying wire loop in a uniform magnetic field

perpendicular to the flux density vector ~ B of the uniform magnetic field. The wire loop carries
an electrical current~i of the direction as shown. Let θ be the angle between ~
B and the plane of
the loop. Also, let the lengths of the loop rectangle be jN1 N2 j ¼ jN3 N4 j ¼ L and
jN2 N3 j ¼ jN4 N1 j ¼ b.
This loop experiences a torque, computed in the following way.
The magnetic forces, exerted by the field on the sides of the wire loop are

f 23 ¼ b~i 23  ~

Side N2 N3 ~ B

Side N4 N1 ~
f 41 ¼ b~i 41  ~
B (6.3.10)

The force ~f 23 is shown, while the force ~f 41 is omitted for the clarity of the illustration.
These forces have the same magnitude – j ~ f j ¼ j~
23 f j ¼ bi B sin θ – but opposite direc-
41
tions according to the right-hand rule. Additionally, these force act along the same line – the
loop axis. Thus, they cancel each other and do not generate a torque.
For the forces acting on two other sides of the loop the situation is different.

f 12 ¼ L~i 12  ~

Side N1 N2 ~ B

Side N3 N4 ~
f 34 ¼ L~i 34  ~
B (6.3.11)

Since the sides N1 N2 and N3 N4 of the loop are parallel to the loop axis, and the axis is
perpendicular to the magnetic field flux density, the vectors~i 12 and~i 34 are perpendicular to
B. Thus, the forces ~
~ f and ~
12 f have the same magnitude: j ~
34 f j ¼ j~
12 f j ¼ LBi , and
34
 6.3 Fundamentals of Electromechanical Systems 429

opposite directions (as shown in Figure 6.3.3). Since these forces act along different lines,
they produce a torque about the loop axis, making the wire loop rotate around its axis. This
torque is derived as

b b
T ¼ f12 þ f34 ¼ bLBi (6.3.12)
2 2

When a conductor moves in a magnetic field in a certain manner, the field induces a
voltage in the conductor. This voltage is sometimes called the electromotive force (emf ). The
phenomenon of voltage generation (electromagnetic induction) is expressed by a fundamental
law of electromagnetism – Faraday’s law of induction. It states that any change in the
magnetic environment of a conductor will cause an emf to be induced in that conductor.
The induced emf is proportional to the negative rate of change of the magnetic flux.
A change in a magnetic environment can be produced in many different ways, such as, for
example, changing the position of a conductor in relation to the magnetic field, or varying the
strength of the field. The systems discussed further in this chapter employ conductors moving
in the uniform magnetic field (~ B ¼ const).
Consider a conductor of length L moving perpendicularly to the magnetic flux lines of a
uniform field of a permanent magnet. Vectors of the magnetic field flux density ~ B and
velocity of the conductor ~ v are as indicated in Figure 6.3.4. According to Lorentz’s force
law, a magnetic field exerts a force on any charged particle moving inside that field. Under the
influence of this magnetic force (Eq. (6.3.1)) the negatively charged particles inside the
conductor move toward the e end of the conductor, while the positive charges congregate
at the opposite, eþ , end. Movement of the charged particles under the influence of the
magnetic force constitutes the electric current~i inside the conductor. The resulting potential
difference between the two ends of the conductor is the induced emf. The direction of the
induced current ~i is determined by Lenz’s law, which states that the magnetic field, corres-
ponding to the induced current, opposes the change in the magnetic flux that induced the
current.

Figure 6.3.4 Direction of the induced current according to Lenz’s law

430 6 Combined Systems and System Modeling Techniques

Since the voltage is defined as the work needed to move a unit charge between two points
(Eq. (4.1.3)), the induced emf εb is the work performed by the magnetic force moving a unit
charge along the length of the conductor:

fL qvBL
εb ¼ ¼ ¼ vBL (6.3.13)
q q

The direction of the induced current is determined by the right-hand rule as shown in
Figure 6.3.4: when index finger points in the direction of the magnetic field flux density ~
B,
and the thumb points in the direction of the conductor’s velocity ~
v, the middle finger will
~
indicate the direction of the induced current i.
To understand how the electromagnetic laws described by Eqs. (6.3.9) and (6.3.13) are
used to express the coupling between mechanical and electrical subsystems of a compound
dynamic system, consider the following example. Let the conductor depicted in Figure 6.3.4
have a non-negligible mass m. According to Eq. (4.1.6) the electrical power due to the
induced emf εb and its corresponding electrical current ~i is

P E ¼ εb i (6.3.14)

Substituting the expression for the induced emf (Eq. (6.3.13)) into Eq. (6.3.14), we obtain

PE ¼ vBLi (6.3.15)

The power applied to the mass m, making it move, is

PM ¼ fv ¼ BLiv (6.3.16)

where f is the magnetic force, described by Eq. (6.3.9).

Assuming that the energy losses due to the conductor resistance and friction are negligible,
the power acting on the mass equals the generated electrical power:

PE ¼ P M or εb i ¼ fv (6.3.17)

For the known representation of the magnetic force (Eq. (6.3.9)), Eq. (6.3.17) derives the
expression for the induced emf (Eq. (6.3.13)).
Also, according to Newton second law, the relation between electrical and mechanical
parameters of this dynamic system is

ma ¼ f ¼ BLi (6.3.18)

where a ¼ dv=dt is the acceleration of the mass m.

6.3.2 D’Arsonval Meter

The simplest electromechanical system is the D’Arsonval meter or galvanometer, illustrated in
Figure 6.3.5. In the figure, the numbers correspond to the following galvanometer elements:
 6.3 Fundamentals of Electromechanical Systems 431

(1) the moving coil of wire of the size l × b as shown, and consisting of n loops (often called
the armature); (2) the assembly of a torsional spring (hairspring) and jewel bearing; (3) the
pointer, (4) the scale; (5) the horseshoe magnet with the polarity as indicated; and (6) the soft-
iron armature core.
A galvanometer is a current-measuring instrument used in popular analog devices such as
ammeters, voltmeters, and ohmmeters. While a detailed discussion on these devices belongs
to the field of electrical engineering, understanding of D’Arsonval movement or, as it is
sometimes called, permanent-magnet moving-coil movement is crucial for comprehending
DC motor operation.
D’Arsonval movement is based on a stationary permanent magnet and a movable coil of
wire that is connected to a source of current to be measured. The coil (armature) is wound
around an iron core, forming an inductor that is positioned within a magnetic field, created
by the permanent magnet. The iron core of the inductor concentrates the magnetic field, thus
strengthening its effect on the current-carrying armature.
Connecting the coil to a voltage source vin closes the coil circuit and produces a current~i that
needs to be measured. The magnetic field of the flux density ~ B exerts a torque on the current-
carrying coil, as discussed earlier in this section (see Figure 6.3.3). For a single wire loop, the
generated torque is expressed by Eq. (6.3.12), while for an armature consisting of n loops it
becomes

T ¼ nblBi (6.3.19)

where, as shown in Figure 6.3.5, i is the current to be measured, B is the flux density of the
magnetic field, b and l are the dimensions of a single loop, and n is the number of loops in the coil.

4 5 6 7
3 8
2 (4)
(4)
9
1

10

(3) θ
0

(3)
(2)
(5)
T

(1) B

(1)
l
b
N S
(2)
i i
+ – (6)
vin (2)
(a) (b)

Figure 6.3.5 D’Arsonval meter: (a) construction of the device; (b) moving coil component (see text for
further details)
432 6 Combined Systems and System Modeling Techniques

The armature is connected to a pointer, extended out to a scale. Since the angular
displacement of the pointer is related to the generated torque, and, in turn, to the amount
of current through the armature, the scale is marked to show this applied current.
Hairsprings and bearings resist the movement of the coil, and return it to its neutral (initial)
position when there is no current. Additionally, when the generated torque is balanced by the
restoring forces of the spring-bearing assembly, the pointer stops moving and the reading can
be obtained.
Let the stiffness and damping coefficients of the spring-bearing assembly be kT and
cT , respectively. Let also the inertia of the inductor (core and coil assembly) be I, and the
angular displacement of the pointer be θ. Then, the equation of rotational motion of the
inductor is

d2 θ dθ
I ¼ cT  kT θ þ T (6.3.20)
dt2 dt

Substituting the expression for the generated torque (Eq. (6.3.19)) into Eq. (6.3.20), we
obtain the relation between the measured current and the angular displacement of the
pointer:

d2 θ dθ
I 2
þ cT þ kT θ ¼ nblBi (6.3.21)
dt dt

As discussed earlier in this section, when a conductor (the armature) rotates in the
magnetic field, an emf is induced, proportional to the conductor’s linear velocity and length
(see Eq. (6.3.13)). Recalling the relation between linear and angular velocities derived in
Chapter 3, the linear velocity of the armature is

b dθ
v¼ (6.3.22)
2 dt

The dimension of the armature that is perpendicular to the magnetic flux density vector is
2nl. Hence, the induced emf is

dθ
εb ¼ vBð2lÞ ¼ nblB (6.3.23)
dt

Let the inductance and resistance of the armature solenoid be L and R, respectively.
Then, remembering that the induced emf has a pol- R L
arity that opposes the change in magnetic flux
responsible for the induced voltage, the armature + +
circuit can be represented as shown in Figure 6.3.6. ε
vin b
Applying Kirchhoff’s voltage law to this circuit,
we obtain – –

di Figure 6.3.6 Electrical subsystem of

vin  εb  Ri  L ¼0 (6.3.24) a galvanometer
dt
 6.4 Direct Current Motors 433

Substituting the derived expression for the induced emf (Eq. (6.3.23)), we obtain the
relationship between the voltage input and the current to be measured:

di dθ
L þ Ri þ nblB ¼ vin (6.3.25)
dt dt

Hence, the model of a galvanometer consists of two equations – Eqs. (6.3.21) and (6.3.25).
When the resistive forces due to the spring–damper assembly and the induced emf become
di d2 θ
equal to the generated torque, the pointer stops moving. That means that ¼ 0, 2 ¼ 0,
dt dt
dθ
and ¼ 0, which transforms Eqs. (6.3.21) and (6.3.25) into
dt
vin
Ri ¼ vin ) i ¼
R (6.3.26)
kT θ ¼ nblBi

Hence, the pointer displacement can be related to the measured current as

nblB
θ¼ i (6.3.27)
kT

and to the input voltage as

nblB
θ¼ vin (6.3.28)
RkT

Equations (6.3.27) and (6.3.28) yield the important parameters of a galvanometer: the
nblB nblB
current sensitivity – the deflection per unit current, and the voltage sensitivity – the
kT RkT
deflection per unit voltage.
The galvanometer is a reliable, sensitive, and accurate measuring device, its accuracy
typically being approximately 2–5% of its full-scale deflection. In all three types of analog
devices – voltmeters, ammeters, and ohmmeters – the galvanometer element reacts to the
amount of current passing through its armature. The principal difference between these meters
lies in the manner in which the galvanometer is connected in the device’s electrical circuit.

6.4 Direct Current Motors

6.4.1 Fundamentals of DC-Motor Operations
The purpose of any electric motor, whether of DC or alternating current (AC) type, is to
convert electric energy into mechanical movement represented by the rotation of the motor’s
434 6 Combined Systems and System Modeling Techniques

shaft. A DC motor is often called a “mechanical workhorse” since a wide variety of

mechanical devices depend on it for the power to move. Easy controllability of the DC
motor rotation speed, high torque capabilities, sturdy construction, and relatively low cost
make it a ubiquitous component of a wide variety of machines, for example, cranes and
winches, cars, and even missile launchers.
The operation of a DC motor is based on D’Arsonval movement, as discussed in the
previous section. The construction of a brushed DC motor is illustrated in Figure 6.4.1, in
which the numbers denote the following elements: (1) permanent magnets (stator) bonded to
the steel enclosure, (2) armature winding, (3) the rotor, (4) the commutator, (5) the brushes,
(6) the shaft, (7) the support bearing, and (8) the steel enclosure.
As discussed earlier, for D’Arsonval movement two major components are needed: a
stationary magnet and a current-carrying movable inductor, which is a coil of wire wound
around an iron core.
The inductor component of a DC motor is an armature, which consists of the multiple coils
of wire (armature winding) attached to a rotor – an iron core that is mounted on a shaft
supported by bearings (see Figure 6.4.1). The rotating armature generates the sought
mechanical power.
The stationary magnetic field around the inductor (armature) is provided by a stator that is
either a permanent magnet (see Figure 6.4.1), or an electromagnet. The stator is permanently
attached to a steel enclosure.
Recalling the discussion on a single current-carrying wire loop rotating in a uniform
magnetic field (see Figure 6.3.3), it can be easily seen that as soon as this coil rotates 180°,
the direction of the generated torque will be reversed as illustrated in Figure 6.4.2.
In Figure 6.4.2 ~f 12 and ~
f 34 denote the forces exerted by the magnetic field ~
B on the N1 N2
and N3 N4 sides of the coil, respectively, and T indicates the resulting torque.
In general, this torque reversal is unacceptable; hence, some measures need to be devised to
provide continuous rotation. Also, permanent electrical contact between the rotating

(5) (3)
(5) (3) (8) (4)
(4)
(2)
(2)

S N
(6) (7)
vin (1) (1)
(1)
vin
(5)

(a) (b)

Figure 6.4.1 DC motor: description and operation: (a) cutaway view, (b) cross-section (see text for further
details)
 6.4 Direct Current Motors 435

i N2 N3
i
N2
N3 f1–2 i
B B
f1–2
i
B
i
f3–4
B
f3–4 N1 N4
i

N4 T T
N1

– vin + + vin –

(a) (b)

Figure 6.4.2 Current-carrying wire loop rotating in a uniform magnetic field: (a) initial position of a coil
and (b) position after the coil has rotated 180°

armature and the motor’s power supply must be ensured for the duration of motor oper-
ation. For these purposes, commutator and brush components are devised. A commutator is
a segmented sleeve, typically made of copper, mounted on the rotating shaft of a DC motor.
The current is supplied to the commutator segments via brushes – two spring-loaded carbon
sticks. As the motor’s shaft turns, the commutator turns with it and the brushes slide over the
commutator segments, supplying electrical energy to the armature via the commutator.
The simplest commutator consists of only two segments. It reverses the direction of the
current flowing through the armature every half-rotation (180°), thus providing for a
continuous rotation in one direction. While perfectly workable, this type of motor construc-
tion is hardly practical due to the inherent problems of a zero-torque position and brush
short-circuiting.
The rotation of a single wire loop connected to a two-segment commutator with brushes
inside the magnetic field of a permanent magnet with indicated polarity is shown in Figure
6.4.3. In the figure, the label (1) denotes the two-segment commutator and (2) the brushes; b
and l are the dimensions of the coil, ~B is the magnetic flux density, ~
f is the exerted magnetic
force, α is the rotation angle,~
n is the normal to the coil surface, and θ is the angle between the
normal and the magnetic flux density.
Recalling the expressions for the exerted magnetic force (Eq. (6.3.11)), the torque is
computed as follows:

b
T ¼ 2f ¼ ðlBiÞb sin θ (6.4.1)
2
436 6 Combined Systems and System Modeling Techniques

f
f
S α b
B α Coil
B
(2)
l θ
(2) b
n —
2 f
(1) f

vin N

(a) (b)

Figure 6.4.3 Rotation of a single wire loop connected to a two-segment commutator: (a) rotating coil; (b)
geometry related to the computation of the torque (see text for further details)

Since the area of this coil is A ¼ bl, the expression for the generated torque can be rewritten as

T ¼ BiA sin θ (6.4.2)

From Eq. (6.4.2) it is obvious that the maximum torque is generated when the normal to
the coil is perpendicular to the vector of magnetic flux density (T max ¼ Bi A), and the
minimum torque (Tmin ¼ 0) occurs when this normal is parallel to ~
B. If such a zero-torque
position of the motor’s armature occurs at powering up, the motor will not be able to start –
since no torque is generated, the rotor will stay motionless despite the applied voltage. It is
less detrimental when the rotation is already in progress since the rotor’s angular momentum
will help it pass the dangerous position without stopping.
Commutator and brush positions at maximum and minimum torque generation are
illustrated in Figure 6.4.4.
Another weakness of a DC motor with the two-segment commutator is the brush short-
circuiting. At a zero-torque position, as shown in Figures 6.4.4(b) and (d), the brushes touch both
commutator segments simultaneously, causing a short circuit. Besides wasting energy, these
short circuits (one per brush) may result in commutator and brush damage due to overheating
and the sparks that result from the short-circuiting, as well as the armature damage.
To avoid these torque ripples and short circuits, modern DC motors use multi-segment
commutators if mechanical commutation is desired, or a brushless construction with elec-
tronic commutation. Spatial distribution of multiple loops of the armature winding is also
helpful in alleviating the mentioned weaknesses.

6.4.2 DC-Motor Modeling

As stated above, there are two ways to manage the performance of a DC motor: either by
varying the electrical current through the armature assembly, or by changing the magnetic field
flux of the stator. Thus, a DC motor can be either armature-controlled or field-controlled.
 6.4 Direct Current Motors 437

B
T B
f1 a1
a1

f2
a2 a2

1 i + i +
vin 1 2 vin
2 – –

(a) (b)

B
T B
f2 a2
a2

f1
a1 a1

2 i + i +
vin 2 1 vin
1 – –

(c) (d)

Figure 6.4.4 Rotation of a single wire loop connected to a two-segment commutator – operation: (a)
maximum-torque position, θ ¼ 90°; (b) zero-torque position, brush short-circuiting, θ ¼ 180° (counter-
clockwise rotation direction is assumed); (c) maximum-torque position, reversal of current direction,
θ ¼ 270°; and (d) zero-torque position, brush short-circuiting, θ ¼ 0°

The electrical subsystem of a typical DC motor, as shown in Figure 6.4.5, is represented by two
circuits: an armature circuit, responsible for generating an electrical current through the armature
assembly, and a field circuit, responsible for providing a magnetic field of the stator. Two power
supplies are thus required – one for the armature circuit, and one for the field circuit.
For an armature-controlled motor, the field circuit may be substituted with a permanent
magnet, leaving this system with a single voltage input. Obviously, for a field-controlled
motor, both circuits must be present.
The mechanical subsystem of a DC motor is represented by the inertial element J, which
embodies the moment of inertia of the armature about the rotation axis, and damping
element cb that represents the bearings supporting the shaft of the rotating armature (see
Figure 6.4.1).
A torque, generated by a DC motor, is proportional to a current in either the armature or
field circuit, depending on the employed control method. The relations between the torque
and these currents are summarized in Table 6.4.1, where T refers to the produced torque, if is
438 6 Combined Systems and System Modeling Techniques

Table 6.4.1 Torque–current relations of a DC motor

Armature-controlled DC motor Field-controlled DC motor

if ¼ const ia ¼ const
T ¼ Kia T ¼ Kif

if
Field
Rf circuit
Lf

Armature circuit

+ La
Ra +
vin εb J

– ia – Cb
dθ
T =ω
dt
Electrical subsystem Mechanical subsystem

Figure 6.4.5 DC motor: electrical and mechanical subsystems

the current through the field circuit, ia is the current through the armature circuit, and K
denotes the motor torque constant.
The expressions for the torque shown in Table 6.4.1 are the coupling relations since they
describe the interaction between the electrical and mechanical subsystems of a DC motor. It
is clearly seen that these coupling expressions contain the variables of both electrical and
mechanical subsystems.

Armature-Controlled DC Motor
Consider an armature-controlled DC motor, where the stator is a permanent magnet, and vin
is the input voltage into the armature circuit. In the context of this control scheme, the
magnetic flux of the stator field is constant, and the generated torque is regulated by
modifying the current through the armature circuit.
 6.4 Direct Current Motors 439

Recalling the procedure for modeling the combined systems described in the Section 6.1
we can derive the governing expressions separately for each subsystem.
For the mechanical subsystem, the free-body diagram of which is shown in Figure 6.4.6,
the derivation is straightforward.
Let θ be the rotation angle of the armature and Tb the T Tb
motion-resisting torque generated in the shaft bearings in
J
response to the angular velocity across the bearings.
Then, according to the Newton’s second law, for a rota- dθ
θ =ω
tional system the equation of motion is dt

J θ€ ¼ T  Tb (6.4.3) Figure 6.4.6 DC motor:

mechanical subsystem
Recalling the expression for the torque exerted by a torsional damper:

Tb ¼ cb θ_ (6.4.4)

the governing equation for the mechanical subsystem is obtained, in terms of angular
displacement (rotation angle):
J θ€ þ cb θ_ ¼ T (6.4.5)
dθ
or in terms of angular velocity ω ¼ :
dt
J ω_ þ cb ω ¼ T (6.4.6)

To derive the governing equations for the electrical subsystem shown in Figure 6.4.7, recall
the discussion about D’Arsonval movement (see Section 6.3.2) and the concept of the
induced emf (or back emf, as it is often called in the context of a DC motor).
Let vin be the voltage input into the armature circuit,
and εb be the back emf, or voltage generated in the
La
current-carrying armature in response to the rotation of Ra
the armature assembly in the magnetic field of the stator; vin ia
εb
Ra and La denote the resistance and inductance of the
circuit elements, respectively, while ia indicates the cur-
rent in the armature circuit. Figure 6.4.7 DC motor:
As previously discussed, the polarity of the generated electrical subsystem
back emf is such that it opposes the change in the mag-
netic flux responsible for the induced voltage. Applying Kirchhoff’s voltage law to the
armature circuit, we obtain

dia
vin  εb  Ra ia  La ¼0 (6.4.7)
dt

As derived earlier (see Eq. (6.3.23)), the back emf is

440 6 Combined Systems and System Modeling Techniques

dθ dθ
εb ¼ nblB ¼ Kb ¼ Kb ω (6.4.8)
dt dt

Equation (6.4.7) is the governing equation for the electrical subsystem, and Eq. (6.4.8) is one of
the coupling equations since it expresses the relation between the electrical parameter εb and the
dθ
mechanical parameter ω ¼ . The second coupling equation is the one that defines the relation
dt
between the produced torque T and the current in the armature circuit ia shown in Table 6.4.1:

T ¼ Kia (6.4.9)

From Eq. (6.3.19) the value of the motor’s torque constant is found as

K ¼ nblB ¼ Kb (6.4.10)

Obviously, the motor’s torque constant and its back emf constant are identical.
Nonetheless, for further derivations we will use K and Kb so that the influence of each of
these constants on the motor performance is more clearly demonstrated.
Thus, the model of an armature-controlled DC motor consists of the following four
equations:
8 €
>
>J θ þ cb θ_ ¼ T
>
< dia
vin  εb ¼ Ra ia þ La
dt (6.4.11)
>
>εb ¼ Kb θ_
>
:
T ¼ Kia

or in terms of angular velocity:

8
>
>J ω_ þ cb ω ¼ T
>
< dia
vin  εb ¼ Ra ia þ La
dt (6.4.12)
>
>εb ¼ K b ω
>
:
T ¼ Kia

This system can be further modeled by any of the methods discussed earlier, such as
transfer function formulations, block diagrams, and state-space representations.
To derive the transfer function formulation and subsequently construct the block dia-
gram, take the Laplace transform of Eq. (6.4.11), assuming zero initial conditions. Let the
voltage vin ðtÞ applied to the armature circuit be the system input, and the motor angular
dθ
velocity ωðtÞ ¼ be the system output.The governing equations in the Laplace domain are
dt
8
>
> ðJs2 þ cb sÞΘðsÞ ¼ TðsÞ
>
<
Vin ðsÞ  Eb ðsÞ ¼ ðRa þ La sÞIa ðsÞ (6.4.13)
>
> Eb ðsÞ ¼ Kb sΘðsÞ
>
:
TðsÞ ¼ KIa ðsÞ
 6.4 Direct Current Motors 441

Table 6.4.2 Armature-controlled DC-motor block diagram components

Governing equation Block diagram component

T(s ) 1 Θ(s)
(Js 2 + cbs)Θ(s) = T(s)
Js 2 + cbs

Vin(s ) + 1 Ia(s )
Vin(s)–Eb(s) = (Ra + Las)Ia(s) Ra + Las
–
Eb(s )

Θ(s ) Eb(s )
Eb(s) = Kbs Θ(s) Kbs

Ia(s ) T(s )
T(s) = KIa(s) K

Figure 6.4.8 DC motor: block diagram

The system transfer function then derives as

L ½ωðtÞ ΩðsÞ sΘðsÞ K

GðsÞ ¼ ¼ ¼ ¼ (6.4.14)
L ½vin ðtÞ Vin ðsÞ Vin ðsÞ ðJs þ cb ÞðRa þ La sÞ þ KKb

As seen from Eq. (6.4.14), the armature-controlled DC motor is a second-order dynamic

system.
Using the derived transfer function (Eq. (6.4.14)), the simplest block diagram for this
system is constructed as shown in Figure 6.4.8.
While this block diagram is suitable for implementation in software for simulation and
further analysis, its expanded form is more suitable for a deeper understanding of the system
behavior and interactions between its subsystems. To construct an expanded block diagram,
the components that correspond to each of the four governing equations (Eq. (6.4.13)) are
built as shown in Table 6.4.2.
Finally, the expanded block diagram is constructed from the above components as
depicted in Figure 6.4.9.
442 6 Combined Systems and System Modeling Techniques

Vin(s) + 1 Ia(s ) T(s ) 1 Θ(s) Ω(s)

K 2 s
– Ra + Las Js + cbs

Eb(s ) Θ(s )
Kbs

Figure 6.4.9 Armature-controlled DC motor: expanded block diagram

To derive the state-space representation of this dynamic system, consider the governing
equations for every individual subsystem since they determine the choice of state variables.
Obviously, coupling equations do not introduce any new state variables since they describe
the relations between parameters of different subsystems that have already been expressed
with the state variables.
The mechanical subsystem is described by Eq. (6.4.5), which is a second-order differential
equation in terms of angular displacement (rotation angle) θðtÞ. Thus, it needs two state
_
variables: x1 ¼ θðtÞ and x2 ¼ θðtÞ.
The electrical subsystem is described by Eq. (6.4.7), which is a first-order differential
equation in terms of the current ia ðtÞ. Hence, it needs a single state variable x3 ¼ ia ðtÞ. The
first two of the selected three state variables
8
<x1 ¼ θðtÞ
_
x ¼ θðtÞ (6.4.15)
: 2
x3 ¼ ia ðtÞ

yield the first state equation:

x_ 1 ¼ x2 (6.4.16)

Rewriting Eq. (6.4.5) in terms of state variables, we obtain

J x_ 2 þ cb x2 ¼ T (6.4.17)

while Eq. (6.4.7) transforms into:

vin  εb ¼ Ra x3 þ La x_ 3 (6.4.18)

Similarly, the coupling relations are expressed. Equation (6.4.9) becomes

T ¼ K x3 (6.4.19)

and Eq. (6.4.8) is

ε b ¼ Kb x 2 (6.4.20)
 6.4 Direct Current Motors 443

From Eqs. (6.4.17) and (6.4.19) the second state equation is derived as

1 cb K
x_ 2 ¼ ðcb x2 þ TÞ ¼  x2 þ x3 (6.4.21)
J J J

The third state equation is derived from Eqs. (6.4.18) and (6.4.20) as

1 Kb Ra 1
x_ 3 ¼ ðεb  Ra x3 þ vin Þ ¼  x2  x3 þ vin (6.4.22)
La La La La

The output equation is

y ¼ θ_ ¼ x2 (6.4.23)

Hence, the state-space representation of an armature-controlled DC motor is

8
>
> x_ 1 ¼ x2
>
>
<x_ 2 ¼  cb x2 þ K x3
State equations J J
>
>
>
> K b Ra 1
:x_ 3 ¼  x2  x3 þ vin ðtÞ
La La La
Output equations yðtÞ ¼ x2 (6.4.24)

If desired, the derived state-space representation can be cast in the matrix form:

_ ¼ AxðtÞ þ BuðtÞ
xðtÞ
yðtÞ ¼ CxðtÞ þ DuðtÞ

where matrices are as follows:

2 3
0 1 0 0 1
6 7 0
c
60  b K 7 B0 C
A3  3 ¼ 6
6 J J 7 B C
7; B3  1 ¼ @ A ; C 1  3 ¼ ð0 1 0 Þ; D ¼ 0
4 Kb 5 1
0   Ra La
La La
0 1
x1
xðtÞ ¼ @x2 A; uðtÞ ¼ vin ðtÞ
x3

Field-Controlled DC Motor
A field-controlled DC motor, shown in Figure 6.4.10, regulates the generated torque by
modifying the magnetic field flux of the stator while keeping the current in the armature
444 6 Combined Systems and System Modeling Techniques

+
vin if Field
Rf circuit
–
Lf

Armature circuit

La
Ra +
J
ia εb
Cb
–
dθ
T =ω
dt

Electrical subsystem Mechanical subsystem

Figure 6.4.10 Field-controlled DC motor: mechanical and electrical subsystems

circuit constant. As depicted, this type of motor has two separate inputs – voltage input vin
into the field circuit and current input into the armature circuit.
The mechanical subsystem of a field-controlled DC motor is identical to that of an
armature-controlled motor, discussed above. Its governing equation is expressed mathem-
atically by Eqs. (6.4.5) and (6.4.6).
The electrical subsystem consists of two circuits as seen in Figure 6.4.10, where the
armature circuit is supplemented by a dedicated control circuit to keep the current ia
constant, counteracting the effect of the back emf εb.
Let vin be the voltage input into the field circuit. No back emf is generated since the current-
carrying stator assembly does not rotate. Rf and Lf denote the resistance and inductance of
the circuit elements, respectively, while if indicates the current in the field circuit.
Then, applying Kirchhoff’s voltage law to the field circuit, the governing equation for the
electrical subsystem is obtained:

dif
vin  Rf if  Lf ¼0 (6.4.25)
dt

Typically, the stator’s magnetic field flux is a nonlinear function of the current in the field
circuit if . According to Eq. (6.3.19) the generated torque is

T ¼ nblBðif Þia ¼ ½nblia Bðif Þ (6.4.26)

Since the current in the armature circuit is kept constant, the parameter ½nblia  is constant,
and the torque produced in the rotating armature assembly is also a nonlinear function of the
 6.4 Direct Current Motors 445

current in the field circuit. For simplicity, an assumption of linear dependence between the
torque and current is made, which yields the following coupling relation:

T ¼ Kif (6.4.27)

where K is the motor torque constant.

Hence, there are only three governing equations for a field-controlled DC motor:
8
€ _
<J θ þ cb θ ¼ T
>
dif
> vin ¼ Rf if þ Lf (6.4.28)
: dt
T ¼ Kif

Taking the Laplace transform of Eq. (6.4.28), we obtain

8
>
<ðJs2 þ cb sÞΘðsÞ ¼ TðsÞ
V ðsÞ ¼ ðRf þ Lf sÞIf ðsÞ (6.4.29)
>
: in
TðsÞ ¼ KIf ðsÞ

Considering the system input to be the voltage input vin into the field circuit, and the output
dθ
to be the motor angular velocity ωðtÞ ¼ , the system transfer function is derived as
dt
L ½ωðtÞ ΩðsÞ sΘðsÞ K
GðsÞ ¼ ¼ ¼ ¼ (6.4.30)
L ½vin ðtÞ Vin ðsÞ Vin ðsÞ ðJs þ cb ÞðRf þ Lf sÞ

As seen from Eq. (6.4.30), the field-controlled DC motor is also a second-order dynamic
system.
The construction of the expanded block diagram for this dynamic system is done using the
procedure similar to that described for the armature-controlled motor. The individual
components of the sought block diagram are presented in Table 6.4.3, and the complete
schematic is shown in Figure 6.4.11.
It is easy to show that the block diagram in Figure 6.4.11 yields the system transfer function
expressed by Eq. (6.4.30). Note that unlike an armature-controlled motor a field-controlled
one does not have a feedback loop since a back emf is not generated in the field circuit.
The mechanical subsystem of a field-controlled DC motor is a second-order differential
equation in terms of angular displacement (rotation angle) θðtÞ, while its electrical subsystem
is a first-order differential equation in terms of the current if ðtÞ. Thus, the choice of the three
state variables needed is
8
<x1 ¼ θðtÞ
_
x2 ¼ θðtÞ (6.4.31)
:
x3 ¼ if ðtÞ
446 6 Combined Systems and System Modeling Techniques

Table 6.4.3 Field-controlled DC-motor block diagram components

Governing equation Block diagram component

(Js 2 + cbs)Θ(s) = T(s)

T(s ) 1 Θ(s )
Js 2 + cbs

Vin(s ) 1 If (s )
Vin(s) = (Rf + Lf s)If (s)
Rf + Lf s

If (s ) T(s )
T(s) = KIf (s) K

Figure 6.4.11 Field-controlled DC motor: expanded block diagram

Rewriting the governing equations in terms of state variables, we obtain

8
<J x_ 2 þ cb x2 ¼ T
v ¼ Rf x3 þ Lf x_ 3 (6.4.32)
: in
T ¼ Kx3

Equations (6.4.31) and (6.4.32) together with the output equation (Eq. (6.4.23)) yield the
following state-space representation of a field-controlled DC motor:
8
> x_ 1 ¼ x2
>
>
>
>
< cb K
State equations x_ 2 ¼  x2 þ x3
> J J
>
>
>
> Rf 1
:x_ 3 ¼  x3 þ vin ðtÞ
Lf Lf
Output equations yðtÞ ¼ x2 (6.4.33)

which may be cast in matrix form

_ ¼ AxðtÞ þ BuðtÞ
xðtÞ
yðtÞ ¼ CxðtÞ þ DuðtÞ
 6.4 Direct Current Motors 447

where matrices are as follows:

2 3
0 1 0 0 1
6 7 0
6 0  cb K 7
6 7
J 7 ; B3  1 ¼ @ 0 C
B
A3  3 ¼ 6 J A ; C1  3 ¼ ð0 1 0 Þ; D ¼ 0
6 7 1
4 Rf 5 Lf
0 0 
Lf
0 1
x1
xðtÞ ¼ @x2 A; uðtÞ ¼ vin ðtÞ
x3

DC Motor with a Load

In its numerous applications, a DC motor never runs idle. It performs useful work by rotating
different mechanical devices such as winches, pumps, or fan blades. Any mechanism that rotates
using the torque generated by a DC motor is called load. While the interaction between the
motor and its load may be quite complex, the most typical interaction is that of a load exerting an
additional torque that opposes the motor torque. This simplification applies an assumption that
the shaft connecting the motor and the load is rigid, massless, and undamped. In this case, the
load and the motor can be lumped together into a single rigid body, the inertia of which is the
sum of inertias of the motor and the load. While the motor and the load rotate with the same
angular velocity, the load torque acts in the opposite direction to the motor torque.
In some cases, the load torque and motor torque may have the same direction. For
example, consider a DC motor used to rotate wheels of a cart. When the cart is moving
downhill, the torque generated by gravity makes the wheels descend; thus, the load torque is
assisting the motor. If the cart moves uphill, the torque due to gravity is still trying to make
cart wheels descend; hence, the load torque resists the motion and opposes the motor.
The presence of a load changes the mechanical subsystem
T TL Tb
as shown in Figure 6.4.12. As before, θ is the rotation angle
of the armature assembly, and Tb is the motion-resisting J
torque generated in the shaft bearings in response to the
dθ
angular velocity. θ =ω
dt
Let TL denote the load torque. Then, the governing equa-
tion for this subsystem becomes Figure 6.4.12 DC motor
with a load: mechanical
J θ€ ¼ T  Tb  TL (6.4.34) subsystem

or
J θ€ þ cb θ_ ¼ T  TL (6.4.35)
448 6 Combined Systems and System Modeling Techniques

The governing equation for the electrical subsystem and the coupling relations stay unchanged.
Therefore, the set of equations that describes an armature-controlled motor with a load
becomes
8 €
>
>J θ þ cb θ_ ¼ T  TL
>
< dia
vin  εb ¼ Ra ia þ La
dt (6.4.36)
>
> _
>
:ε b ¼ K b θ
T ¼ Kia

The load torque depends on the physical characteristics of the load, so in the context of
DC-motor performance it is considered an independent input.
The block diagram component that corresponds to the first line of Eq. (6.4.36) is then
modified as shown in Figure 6.4.13.
Then, the expanded block diagram of an armature-controlled DC motor with a load is
illustrated in Figure 6.4.14.
This block diagram can be reduced as shown in Figure 6.4.15, where two transfer functions:

T(s) + 1 Θ(s)
(Js 2 + cbs)Θ(s) = T(s) – TL(s)
– Js 2 + cbs

TL(s)
(a) (b)

Figure 6.4.13 DC motor with a load: (a) equation in the Laplace domain; (b) block diagram component

TL(s )

Vin(s) + 1 Ia(s ) T(s ) + – 1 Θ(s) Ω(s )

K s
Ra + Las Js 2 + cbs
–

Eb(s ) Θ(s )
Kbs

Figure 6.4.14 Armature-controlled DC motor with a load: block diagram

Vin(s)
Gvin
Ω(s)
TL(s)
GTL

DC motor

Figure 6.4.15 Armature-controlled DC motor with a load: reduced block diagram

 6.4 Direct Current Motors 449

ΩðsÞ
GVin ðsÞ ¼
Vin ðsÞ

and
ΩðsÞ
GTL ðsÞ ¼
TL ðsÞ

are clearly distinguishable.

From the diagram in Figure 6.4.15, the transfer functions are derived as follows:

K s
ðVin ðsÞ  Kb s ΘðsÞÞ  TL ðsÞ ¼ ΩðsÞ
Ra þ La s Js þ cb s
2

K 1 KKb
Vin ðsÞ  TL ðsÞ ¼ 1þ ΩðsÞ
ðRa þ La sÞðJs þ cb Þ Js þ cb Ra þ L a s
ΩðsÞ K
GVin ðsÞ ¼ ¼
Vin ðsÞ ðRa þ La sÞðJs þ cb Þ þ KKb
(6.4.37)
ΩðsÞ Ra þ La s
GTL ðsÞ ¼ ¼
TL ðsÞ ðRa þ La sÞðJs þ cb Þ þ KKb

Sometimes the current in the armature circuit is considered to be the second output of this
dynamic system (in addition to the angular velocity). Then, the additional two transfer
functions are derived as
8
> 1
<ðVin ðsÞ  Kb sΘðsÞÞ ¼ Ia ðsÞ
Ra þ La s
> s
:ðIa ðsÞK  TL ðsÞÞ ¼ ΩðsÞ
Js2 þ cb s

Substituting the value for ΩðsÞ ¼ sΘðsÞ of the second equation into the first one, we obtain

Ia ðsÞK  TL ðsÞ
Vin ðsÞ  Kb ¼ ðRa þ La sÞIa ðsÞ
Js þ cb
Kb ðRa þ La sÞðJs þ cb Þ þ KKb
Vin ðsÞ þ TL ðsÞ ¼ Ia ðsÞ
Js þ cb Js þ cb

which derives the sought transfer functions:

Ia ðsÞ Js þ cb
GVin Ia ðsÞ ¼ ¼
Vin ðsÞ ðRa þ La sÞðJs þ cb Þ þ KKb
Ia ðsÞ Kb
GTL Ia ðsÞ ¼ ¼ (6.4.38)
TL ðsÞ ðRa þ La sÞðJs þ cb Þ þ KKb
450 6 Combined Systems and System Modeling Techniques

The selection of state variables for the case of lumped motor–load inertias is not influenced
by the addition of the load. The state-space representation of this dynamic system is obtained
as
8
> x_ 1 ¼ x2
>
>
>
>
<_ cb K
x 2 ¼  x2 þ x3  1 TL ðtÞ
State equations J J J
>
>
>
>
>
:x_ 3 ¼  Kb x2  Ra x3 þ vin ðtÞ
La La 1La
Output equations yðtÞ ¼ x2 (6.4.39)

and, in the matrix form:

_ ¼ AxðtÞ þ BuðtÞ
xðtÞ
yðtÞ ¼ CxðtÞ þ DuðtÞ

where matrices are

2 3
0 1 0 0 1
6 7 0 0
60  cb K 7 B0  C
1
6 J 7 B C
A3  3 ¼ 6 J 7 ; B3  2 ¼ B J C ; C 1  3 ¼ ð0 1 0 Þ; D ¼ 0
6 7 @1 A
4 Kb R 5 0
0   a
La
La La
0 1
x1 ¼ θðtÞ 
@ _ A v ðt Þ
xðtÞ ¼ x2 ¼ θðtÞ ; uðtÞ ¼ in
TL ðtÞ
x3 ¼ ia ðtÞ

Similarly, for the field-controlled DC motor with a load, the governing equations are
8
€ _
<J θ þ cb θ ¼ T  TL
>
dif
> vin ¼ Rf if þ Lf (6.4.40)
: dt
T ¼ Kif

The modified block diagram is depicted in Figure 6.4.16.

Figure 6.4.16 Field-controlled DC motor with a load: block diagram

 6.4 Direct Current Motors 451

The transfer functions are derived as

ΩðsÞ K
GVin ðsÞ ¼ ¼
Vin ðsÞ ðRf þ Lf sÞðJs þ cb Þ
(6.4.41)
ΩðsÞ 1
GTL ðsÞ ¼ ¼
TL ðsÞ Js þ cb

If the current in the field circuit is considered an additional output, the corresponding
transfer function is

If ðsÞ 1
GVin If ðsÞ ¼ ¼ (6.4.42)
Vin ðsÞ Rf þ Lf s

Note that due to the absence of a feedback loop (unlike the armature-controlled
motor the field-controlled motor does not have a back emf in the field circuit) the
load torque has no influence on the current in the field circuit, so GTL If ðsÞ does not
exist.
The state-space representation of a field-controlled DC motor with a load is as follows:
8
>
> x_ 1 ¼ x2
>
>
>
<x_ ¼  cb x þ K x  1 T ðtÞ
2 2 3 L
State equations J J J
>
>
>
> Rf
>
:x_ 3 ¼  x3 þ 1 vin ðtÞ
Lf Lf
Output equations yðtÞ ¼ x2 (6.4.43)

and, in matrix form:

_ ¼ AxðtÞ þ BuðtÞ
xðtÞ
yðtÞ ¼ CxðtÞ þ DuðtÞ

where the matrices are as follows:

2 3
1 0 0 0 1
6 7 0 0
6 cb 7 B 1C
60  K 7 B0  C
A3  3¼6 J J 7 ; B3  2 ¼ B J C; C 1  3 ¼ ð 0 1 0 Þ ; D ¼ 0
6 7 @1 A
4 Rf 5 0
0 0  Lf
Lf
0 1
x1 ¼ θðtÞ 
v ðtÞ
xðtÞ ¼ @ x2 ¼ θðtÞ
_ A; uðtÞ ¼ in
TL ðtÞ
x3 ¼ ia ðtÞ
452 6 Combined Systems and System Modeling Techniques

Example 6.4.1
Consider an armature-controlled DC motor with a load shown in Figure 6.4.17.
In this assembly, the load is connected to the motor via a flexible damped shaft, with torsional
stiffness kb and torsional damping cb . The load is supported with the bearings that have a damping
coefficient cL . The inertia of the motor is J, while the inertia of the load is JL . The stator of the given
DC motor is represented by a permanent magnet. Ra and La denote the resistance and inductance
of the circuit elements respectively, while ia indicates the current in the armature circuit. The
motor torque constant is K, and the back emf constant is Kb . The voltage input into the armature
circuit is denoted vin ðtÞ.
Considering that the system inputs are vin ðtÞ and the load torque TL ðtÞ, and the system outputs
are the angular velocity of the motor ω and the angular velocity of the load ωL , derive:

(a) the governing equations of this dynamic system

(b) a block diagram
(c) a state-space representation

Solution
As seen in Figure 6.4.17, the mechanical subsystem of this dynamic system is represented with two
inertia elements instead of one: the motor and the load cannot be lumped together into a single
element because of the flexible damped shaft that connects them. Thus, the mechanical subsystem
has two degrees of freedom and needs to be described by two differential equations.
The derivation of the governing equations for the mechanical subsystem starts with drawing a
free-body diagram for each inertia element, and indicating all the acting torques. The free-body
diagrams are illustrated in Figure 6.4.18. Note, that both the motor and the load rotate in the same
direction, while the load torque opposes the motion as discussed earlier.
The inertial element J is subject to three acting torques: TðtÞ – the generated motor torque; Tk –
the restoring torque due to the shaft flexibility that is modeled as a torsional spring; and Tb – the
resisting torque due to the shaft torsional damping arising in response to angular velocity. The
torque Tk works against the angular displacement θðtÞ, while the torque Tb opposes the angular
velocity ωðtÞ ¼ θ._
Assuming that the angular displacement of the motor is greater than that of the load, i.e.
θðtÞ > θL ðtÞ, the expressions for Tk and Tb are (see Chapter 3)

cb
La cL
+ Ra + JL
J
vin εb kb
– ia – dθ = ω d θL
T
dt
TL ωL =
dt
Electrical subsystem Mechanical subsystem

Figure 6.4.17 DC motor with a load assembly for Example 6.4.1

 6.4 Direct Current Motors 453

T Tk Tb Tk Tb TL Tc

J JL

d θL
dθ ωL =
θ =ω θL
dt dt
(a) (b)

Figure 6.4.18 Free-body diagrams of (a) the motor and (b) the load

Tk ¼ kb ðθ  θL Þ
Tb ¼ cb ðθ_  θ_ L Þ

Applying Newton’s second law for a rotational system, the equation of motion for the motor is

J θ€ ¼ TðtÞ  Tk  Tb ¼ TðtÞ  kb ðθ  θL Þ  cb ðθ_  θ_ L Þ (1)

Similarly, the governing equation for the load is derived as

JL θ€L ¼ TL ðtÞ þ Tk þ Tb  Tc ¼ TL ðtÞ þ kb ðθ  θL Þ þ cb ðθ_  θ_ L Þ  cL θ_ L (2)

Equations (1) and (2) represent the governing equations for the mechanical subsystem.
The governing equation for the electrical subsystem, the mechanism of generating the back emf
in response to the rotation of the armature assembly, and the coupling relations are the same as for
the armature-controlled DC motor:

dia
vin  εb ¼ Ra ia þ La
dt
dθ (3)
εb ¼ K b
dt
T ¼ Kia

Hence, the governing equations for this dynamic system are

8
>
> J θ€ þ kb ðθ  θL Þ þ cb ðθ_  θ_ L Þ ¼ TðtÞ
>
>
>
>
>
> JL θ€L  kb ðθ  θL Þ  cb ðθ_  θ_ L Þ þ cL θ_ L ¼ TL ðtÞ
>
<
dia
> vin  εb ¼ Ra ia þ La (4)
>
> dt
>
> dθ
>
> εb ¼ K b
>
>
: dt
T ¼ Kia

To construct a block diagram, take the Laplace transform of Eq. (4) considering zero initial
conditions:
454 6 Combined Systems and System Modeling Techniques

8
>
> ðJs2 þ cb s þ kb ÞΘðsÞ  ðcb s þ kb ÞΘL ðsÞ ¼ TðsÞ
>
>
>
<ðJL s2 þ ðcb þ cL Þs þ kb ÞΘL ðsÞ  ðcb s þ kb ÞΘðsÞ ¼ TL ðsÞ
Vin ðsÞ  Eb ðsÞ ¼ ðRa þ La sÞIa ðsÞ (5)
>
>
>
>
:Eb ðsÞ ¼ Kb sΘðsÞ
>
TðsÞ ¼ KIa ðsÞ

The last three equations in the set of Eq. (5) are easy to transform into block diagram
components, as for the armature-controlled DC motor discussed above, and the resulting com-
ponents are shown in Table 6.4.2.
The construction of the block diagram components that represent the relations between the
torques TðsÞ and TL ðsÞ and the rotation speeds ΘðsÞ and ΘL ðsÞ requires a number of algebraic
transformations of the first two equations of the set of Eq. (5) to obtain the expressions in the form

ΘðsÞ ¼ G1 ðsÞTðsÞ þ G2 ðsÞTL ðsÞ (6)

and
ΘL ðsÞ ¼ G3 ðsÞTðsÞ þ G4 ðsÞTL ðsÞ (7)

To simplify the derivation, let the following be denoted as

Js2 þ cb s þ kb ¼ g1
JL s2 þ ðcb þ cL Þs þ kb ¼ g2
cb s þ kb ¼ g3

Then, the first two equations of Eq. (5) are rewritten as

g1 ΘðsÞ  g3 ΘL ðsÞ ¼ TðsÞ (8)

g2 ΘL ðsÞ  g3 ΘðsÞ ¼ TL ðsÞ (9)

From Eq. (9)

1 g3
ΘL ðsÞ ¼  TL ðsÞ þ ΘðsÞ (10)
g2 g2

Plugging Eq. (10) into Eq. (8), we obtain


1 g3
g1 ΘðsÞ  g3  TL ðsÞ þ ΘðsÞ ¼ TðsÞ
g2 g2

which yields the desired form of Eq. (6):

g2 g3
ΘðsÞ ¼ TðsÞ  TL ðsÞ (11)
g1 g2  g3 2 g1 g2  g3 2
 6.4 Direct Current Motors 455

Eb(s) Θ(s)
Kbs
+ Θ(s) Ω(s)
G1 s
Vin(s) + – 1 Ia(s) T(s) –
K
Ra+Las
+ ΘL(s) ΩL(s)
G3 s
–

G2
TL(s)

G4

Figure 6.4.19 DC motor with a load assembly: block diagram

In a similar manner, the expression for ΘL ðsÞ is derived:

g3 g1
ΘL ðsÞ ¼ TðsÞ  TL ðsÞ (12)
g1 g2  g3 2 g1 g2  g3 2

The expanded block diagram of this dynamic system is shown in Figure 6.4.19, where
g2 g3 g3 g1
G1 ðsÞ ¼ ; G2 ðsÞ ¼ ; G3 ðsÞ ¼ ; G4 ðsÞ ¼
g1 g2  g3 2 g1 g2  g3 2 g1 g2  g3 2 g1 g2  g3 2

This system has four transfer functions:

ΩðsÞ ΩðsÞ ΩL ðsÞ ΩL ðsÞ
GVin ðsÞ ¼ ; GTL ðsÞ ¼ ; GVin L ðsÞ ¼ ; GTL L ðsÞ ¼
Vin ðsÞ TL ðsÞ Vin ðsÞ TL ðsÞ

The derivation of these transfer functions from the constructed block diagram or from the derived
governing equations (Eq. (5)) is straightforward, using the approaches discussed in Sections 6.2.2
and 6.2.3.
The state-space representation of this dynamic system is easily obtained from that derived
earlier by adding two more state variables x4 ¼ θL and x5 ¼ θ_ L .
8
> x_ 1 ¼ x2
>
>
>
> kb cb K kb cb
>
> x_ 2 ¼
x 1  x 2 þ x3 þ x 4 þ x 5
>
>
>
> J J J J J
<
Kb Ra 1
State equations x_ 3
¼  x2  x3 þ vin ðtÞ
>
> La La La
>
>
> x_ 4
¼ x5
>
>
>
>
> cb þ cL
> kb
:x_ 5 cb
¼ x1 þ x2  x4 
kb 1
x5  TL ðtÞ
JL JL JL JL JL

x
Output equations yðtÞ ¼ 2
x5
456 6 Combined Systems and System Modeling Techniques

And, in the matrix form:

_ ¼ AxðtÞ þ BuðtÞ
xðtÞ
yðtÞ ¼ CxðtÞ þ DuðtÞ

where the matrices are

2 3
0 1 0 0 0
6 kb 7 0 1
6 
cb K kb cb 7 0 0
6 7 B0 0 C
6 J J J J J 7 B C
6 7 B1 C
6 0 Kb Ra 7 0 C;
A55 ¼6   0 0 7; B52 ¼ B
BLa C
6 La La 7 B C
6 7 B0 0 C
6 7
6 0 0 0 0 1 7 @ 1A
6 7 0 
4 k
b cb kb cb þ cL 5 JL
0  
JL JL JL JL

0 1 0 0 0
C 25 ¼ ; D¼0
0 0 0 0 1
0 1
x1 ¼ θðtÞ
_
B x2 ¼ θðtÞ C 
B C vin ðtÞ
B C
xðtÞ ¼ B x3 ¼ ia ðtÞ C; uðtÞ ¼
@ x ¼ θ ðtÞ A TL ðtÞ
4 L
x5 ¼ θ_ L ðtÞ

Simplification of a DC-Motor Model

As can be seen from Eq. (6.4.14), the armature-controlled DC motor is a second-order
dynamic system. While the mathematical apparatus for the analysis of such systems is
extensively developed, in many cases reduction of this system to a first-order one is desirable.
Such a reduction is reasonable since, typically, the inductance of the armature circuit
La
greatly exceeds the resistance, i.e. << 1. Then, the expression ðRa þ La sÞ in the denomin-
Ra
ator of the system transfer function (Eq. (6.4.14)) becomes

La s
Ra þ La s ¼ Ra þ 1 ffi Ra
Ra
La s
since is very small.
Ra
Then, the transfer function reduces as follows:
ΩðsÞ K K
GðsÞ ¼ ¼ ¼
Vin ðsÞ ðJs þ cb ÞðRa þ La sÞ þ KKb ðJs þ cb ÞRa þ KKb

K K=ðRa cb þ KKb Þ
¼ ¼ . 
JRa s þ ðRa cb þ KKb Þ JRa ðRa cb þ KKb Þ s þ 1
 6.5 Block Diagrams in the Time Domain 457

Setting coefficients
.
KM ¼ K ðRa cb þ KKb Þ

and
.
τ ¼ JRa ðRa cb þ KKb Þ

presents the transfer function in the familiar form of a first-order system:

ΩðsÞ KM
GðsÞ ¼ ¼ (6.4.44)
Vin ðsÞ τs þ 1

6.5 Block Diagrams in the Time Domain

Transfer function formulations (Section 6.2.2) and block diagrams in the s-domain (Section
6.2.3) are valid for linear time-invariant dynamic systems, for which the governing differen-
tial equations are linear with constant coefficients. These techniques are not applicable to
linear time-variant systems and nonlinear systems, for which the governing differential
equations have time-dependent coefficients and/or nonlinearities. For these systems, state-
space representations (Section 6.2.1) are widely used in modeling and simulation.
In this section, another modeling technique, which is called block diagrams in the time
domain or time-domain block diagrams, is introduced. This technique is applicable to
general dynamic systems, including linear time-invariant systems, linear time-variant sys-
tems, and nonlinear systems. One benefit of learning time-domain block diagrams is to
prepare for the use of the Simulink software in the modeling and simulation of complicated
systems. As will be seen in Section 6.6.1, the structure of time-domain block diagrams and
that of Simulink models are quite similar.
Like an s-domain block diagram, a time-domain block diagram is an assembly of basic
elements or components. Table 6.5.1 lists some basic components for the construction of
time-domain block diagrams. From the table, the signal, summing point, pick-off point, and
constant gain follow the same rules as their counterparts in the s-domain; a differential
operator D was introduced in Section 2.5.1; and integrator can be symbolically expressed by
Ð
dt ¼ D1 . However, three new components (the last three in the table) are not seen in the
s-domain block diagrams: the time-varying gain, multiplier, and function block, which are
used to describe time-variant properties and nonlinear characteristics. A multiplier can be
used to describe the powers of a variable, such as x2 and x3 . A time varying gain gðtÞ or
nonlinear function ϕðtÞ can be used to describe the time-variant or nonlinear property of a
dynamic system.
458 6 Combined Systems and System Modeling Techniques

Table 6.5.1 Basic components of time-domain block diagram

Component Symbol Rule

u v
Signal Variable, input or output

u + w
Summing point w=u–v
–
v

Pick-off point u v=u

v w w=u

du dx
Differentiator D, overdot Du = ⎯ , x· = ⎯
dt dt

y z t
Integrator ∫ z (t ) = ∫ y(τ)dτ
t0

x y
Constant gain k y = kx

x y
Time-varying gain g(t ) y = g (t)x

u w
Multiplier × w = u ×v
v

x1 y
x2 × y = x1× x2× x3
x3

Function block t y
φ(t ) y = φ (t )

Consider a dynamic system with its characteristic variable xðtÞ governed by an n-th-order
differential equation. Denote the system input by rðtÞ. A systematic procedure for construc-
tion of a time-domain block diagram for the system takes the following five steps.

(1) Rewrite the differential equation in the following form:

 6.5 Block Diagrams in the Time Domain 459

Dn x ¼ f ðx; Dx; . . . ; Dn1 x; rÞ (6.5.1)

where f is a function of variable x, its derivatives up to order n − 1, and input r. For a linear
dynamic system, f is a linear function; for a nonlinear system, f is a nonlinear function.
(2) Integrate the left-hand side of Eq. (6.5.1) n times, which gives a part of the block diagram
as shown in Figure 6.5.1(a).
(3) By using the summing point and constant gain in Table 6.5.1, obtain a representation of
function f, which gives another part of the block diagram, as shown in Figure 6.5.1(b).
The representation should take r as an input, and f as an output.
(4) By Dn x ¼ f , assemble the two parts obtained in steps 2 and 3 into a block diagram, as
shown in Figure 6.5.1(c). During the assembly, the basic components in Table 6.5.1 are
used.
(5) Add the system outputs to complete the construction of the block diagram.

The above procedure is illustrated in Examples 6.5.1 through 6.5.4.

Figure 6.5.1 Construction of a time-domain block diagram: (a) integration of Dn x; (b) representation of
function f by x; Dx; . . . ; Dn1 x and r; (c) a completed block diagram
460 6 Combined Systems and System Modeling Techniques

Example 6.5.1
For a spring–mass–damper system, its displacement xðtÞ is governed by

x þ cx_ þ kx ¼ q
m€ (a)

_ be
Let the external force q be the input. Let the displacement x and the damping force ( fd ¼ cx)
the outputs. Build a time-domain block diagram for the system.
Solution
Apply the five-step procedure. In the first step, Eq. (a) is reduced to

1
D2 x ¼ f ≡ ðq  cDx  kxÞ (b)
m

In the second step, D2 x in Eq. (b) is integrated twice, yielding the part in Figure 6.5.2(a). In the
third step, a representation of function f in Eq. (b) is established in Figure 6.5.2(b). In the fourth
step, the two parts obtained in the previous steps are assembled according to D2 x ¼ f . And in the
fifth step, by adding the two designated outputs, the time-domain block diagram for the mechan-
ical system is completed; see Figure 6.5.2(c).

Figure 6.5.2 Time-domain block diagram for the system in Example 6.5.1: (a) integration of D2 x;
(b) representation of f; and (c) completion of the block diagram
 6.5 Block Diagrams in the Time Domain 461

The spring–mass–damper system in Example 6.5.1 is a linear time-invariant system, for

which an s-domain block diagram can be also obtained (see Section 6.2.3). This, of course, is
a trivial case. Now, consider a nonlinear system in the following example, for which the
construction of a block diagram in the s-domain is not possible.

Example 6.5.2
The displacement xðtÞ of a nonlinear mechanical system is governed by

x þ c0 x_ þ c1 x_ 3 þ ðk0 þ ε sin ΩtÞx ¼ q

m€ (a)

Equation (a) indicates that the system experiences a nonlinear damping force, which is
fd ¼ c0 x_ þ c1 x_ 3 , and a time-dependent spring force, which is fs ¼ ðk0 þ ε sin ΩtÞx. Let the external
force q be the input. Let the displacement x and the spring force fs be the outputs. Build a time-
domain block diagram of the system.

D 2x Dx x q + 1 f
³ ³ —
m
–
+
c0Dx + c1(Dx)3
+
(k0 + ε sinΩt )x
(a) (b)

fs
ε sinΩt
+ +
k0
fs
–
+ 1 x·· x·
q
—
m ³ ³ x
–
fd +
c
+

c1 ×

(c)

Figure 6.5.3 A time-domain block diagram in Example 6.5.2: (a) integration of D2 x; (b) representation
of f, and (c) completion of the block diagram
462 6 Combined Systems and System Modeling Techniques

Solution
Equation (a) is first reduced to

1h i
D2 x ¼ f ≡ q  c0 Dx  c1 ðDxÞ3  ðk0 þ ε sin ΩtÞx (b)
m

The derivative D2 x and function f in Eq. (b) are then expressed by the parts in Figures 6.5.3(a)
and (b), respectively. Afterwards, these parts are assembled, and the outputs are added, to give a
complete time-domain block diagram of the system in Figure 6.5.3(c).

The five-step block diagram construction procedure is applicable to systems with multiple
inputs and multiple outputs, which are governed by coupled differential equations. In
handling such a system, one just needs to obtain the two parts for each of the differential
equations separately in steps 1 to 3, and then assemble all the parts in steps 4 and 5 to
complete a block diagram.

Example 6.5.3
A two-input–two-output system is governed by the following coupled differential equations:

x þ 5x þ 0:1x3  y ¼ u
4€ (a)
y_ þ 3y  6xy ¼ v (b)

where u and v are the inputs, and x and y are the outputs. Build a block diagram for the system.
Solution
The differential equations in Eqs. (a) and (b) are rewritten as

Figure 6.5.4 Parts of a block diagram in Example 6.5.3: (a) for Eq. (c); (b) for Eq. (d)
 6.5 Block Diagrams in the Time Domain 463

1
x€ ¼ f1 ≡ ðu  5x  0:1x3 þ yÞ (c)
4
y_ ¼ f2 ≡ v  3y þ 6xy (d)

The two parts of each of the differential equations are drawn in Figure 6.5.4.
Assembling all these parts gives the block diagram of the dynamic system in Figure 6.5.5.
Finally, the construction of a time-domain block diagram for an armature-controlled DC
motor is presented in Example 6.5.4.

Figure 6.5.5 Time-domain block diagram of the system in Example 6.5.3

Example 6.5.4
For an armature-controlled DC motor, the governing equations (6.4.36) are reduced to

1
D2 θ ¼ f1 ≡ ðKia  TL  cb DθÞ (a)
J
1
Dia ¼ f2 ≡ ðvin  Kb Dθ  Ra ia Þ (b)
La
464 6 Combined Systems and System Modeling Techniques

where θ is the rotation angle of the motor; vin and ia are the voltage input and current of the
armature circuit, respectively; T is the motor torque constant; Kb is the back emf constant; and TL
denotes a load torque. As shown in Eq. (6.4.10), K ¼ Kb .
_
Let the inputs of this electromechanical system be vin and TL . Let the output of the system be θ.
By following the five steps listed above, a time-domain block diagram of the motor is obtained in
Figure 6.5.6, where T ¼ Kia is the motor torque.

D 2θ Dθ θ TL – f1
1
³ ³ —
J
+ –
cbD θ
Kia
(a)

Dia ia vin + f2
³
– –
Raia
KbD θ
(b)

TL – 1 θ·· θ
·
—
J ³
+ –
cb

T
K

vin + 1
—
La ³ ia
–
–
Ra

·
θ
Kb

(c)

Figure 6.5.6 A time-domain block diagram for the armature-controlled DC motor in Example 6.5.4:
(a) two parts from Eq. (a); (b) two parts from Eq. (b); and (c) completion of the block diagram

6.6 Modeling and Simulation by Simulink

In this section, the use of Simulink in the modeling and simulation of dynamic systems is
illustrated in examples. Refer to Appendix B.2 on how to use Simulink.
 6.6 Modeling and Simulation by Simulink 465

6.6.1 Simulink Models Based on Time-Domain Block Diagrams

A Simulink model can be easily built based on the concept of block diagrams in the time
domain, as presented in Section 6.5.

Example 6.6.1
Consider the system in Example 6.5.3. Consider zero initial disturbances. Let the inputs of the
system be u ¼ 1:5 sinð10tÞ and v ¼ 0:6. Plot the system outputs x and y for 0 ≤ t ≤ 20.

Solution
According to Figure 6.5.5, a Simulink model of the system is constructed, as shown in Figure
6.6.1. Running the model in simulation gives the system response in Figure 6.6.2. Refer to
Appendix B.2 for how to obtain simulation results (plots) from the Scope block.

Figure 6.6.1 Simulink model of the system in Example 6.6.1

466 6 Combined Systems and System Modeling Techniques

0.1

0.08

0.06
x

0.04

0.02

0 2 4 6 8 10 12 14 16 18 20

0.25

0.2

0.15
y

0.1

0.05

0 2 4 6 8 10 12 14 16 18 20
t

Figure 6.6.2 Response (x and y) of the system in Example 6.6.1

Example 6.6.2
An armature-controlled DC motor, which is governed by Eqs. (a) and (b) in Example 6.5.4, has
the following parameter values:

La ¼ 1:9  103 H; Ra ¼ 0:56 Ω; K ¼ Kb ¼ 0:06 N  m=A

(a)
J ¼ 9  105 kg  m2 ; cb ¼ 1:2  104 N  m  s=rad

Let the applied armature voltage be vin ¼ 10 V. Build a Simulink model of the motor, and plot
_ and the armature current ia ðtÞ of the motor for 0 ≤ t ≤ 0:1 s. In
the rotation speed ωðtÞ ¼ θðtÞ
simulation, the following two cases are considered:
 6.6 Modeling and Simulation by Simulink 467

Case 1: the motor idles with no load torque, TL ¼ 0

Case 2: the motor idles initially and, at t ¼ 0:04 s, a constant load torque, TL ¼ 0:6 Nm is
applied

Solution
By referencing Figure 6.5.6, a Simulink model of this electromechanical system is built; see Figure
6.6.3. The model has two inputs, which are the load torque TL and applied voltage vin ; and two
outputs, which are the rotation speed ω and armature current ia of the motor. With the Simulink
model, the rotation speed and armature current of the motor are plotted against time for the
above-mentioned two cases.
In case 1, as shown in Figure 6.6.4, the rotation speed of the idling motor reaches a
steady-state value of 163.6 rad/s (1562 rpm). The armature current initially goes up as high
as 13.19 A, which is needed to overcome the inertia and damping effects of the motor, and it
eventually settles at 0.33 A.
In case 2, the motor behaves in the same way initially. At time t ¼ 0:04 s when the
load torque TL is applied, the pattern of the system response starts to change. As shown in

Figure 6.6.3 Simulink model of the motor in Example 6.6.2

468 6 Combined Systems and System Modeling Techniques

180
160
140
120
100
ω (rad/s)

80
60
40
20
0
–20
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1

14
12
10
8
ia (A)

6
4
2
0

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
t (s)

Figure 6.6.4 Response of the motor without a load (case 1)

Figure 6.6.5, the motor speed significantly drops and the armature current quickly climbs
up. The steady-state values of the speed and current are 72.0 rad/s (688 rpm) and 10.13 A,
respectively, showing the effect of the applied load. As observed, the applied load toque
reduces the motor speed by 56%.
In reality, the load torque TL is usually time varying and the applied voltage vin takes time to
reach a designated level, say 10 V. Therefore, to maintain a desired motor speed in operation, a
feedback control needs to be implemented. A problem on the feedback control of rotor rotation
speed is presented in Section 9.2.
 6.6 Modeling and Simulation by Simulink 469

180
160
140
120
ω (rad/s)
100
80
60
40
20
0
–20
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1

14
12
10
8
ia (A)

6
4
2
0

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
t (s)

Figure 6.6.5 Response of the motor under a constant load torque (case 2)

6.6.2 Simulink Models with Transfer Functions and State-Space Blocks

Besides building up models from scratch, Simulink can integrate transfer function formula-
tions and state-space representations in modeling and simulation. This is enabled by the
Transfer Fcn block and the State-Space block. See the following two examples.

Example 6.6.3
Consider a time-variant mechanical system with displacement x governed by the following
ordinary differential equation:

x þ c0 x_ þ ðk0  ε sin ΩtÞx ¼ q

m€ (a)
Assign the system parameter values as follows:

m ¼ 2; c0 ¼ 3; c1 ¼ 0:1; k0 ¼ 32; ε ¼ 12; and Ω ¼ 10

470 6 Combined Systems and System Modeling Techniques

Figure 6.6.6 Simulink model of the system in Example 6.6.3

all of which are in SI units. The external force is given as q ¼ 2 sinð4tÞ. Let q and x be the input and
output of the system, respectively. From the time-domain block diagram in Figure 6.5.3(c), a
Simulink model of the system is established; see Figure 6.6.6.
On the other hand, Eq. (a) can be written in the symbolic form

1
x¼ f (b)
mD2 þ c0 D þ k0

where
D ¼ d=dt

and
f ¼ q þ ε sinðΩtÞx (c)

Equation (b) can be expressed by the Transfer Fcn block from the Continuous Library,
which leads to an equivalent Simulink model of the system in Figure 6.6.7.
_
Assume zero initial disturbances: xð0Þ ¼ xð0Þ ¼ 0. Simulation with the original model in
Figure 6.6.6 and the equivalent model in Figure 6.6.7 gives the same displacement plot as
shown in Figure 6.6.8. Refer to Appendix B.2 for how to obtain simulation results (plots) from
the Scope block.
 6.6 Modeling and Simulation by Simulink 471

Figure 6.6.7 Equivalent Simulink model with a Transfer Fcn block

0.2

0.15

0.1

0.05
x

–0.05

–0.1

–0.15

–0.2
0 5 10 15
t

Figure 6.6.8 Displacement of the time-variant system in Example 6.6.3

Example 6.6.4
Consider the armature-controlled DC motor in case 2 of Example 6.6.2. Build a Simulink model
for the motor by a State-Space block, with the motor rotation speed and the motor torque
(T ¼ Kia ) as the system outputs. Also, plot the system outputs for 0 ≤ t ≤ 0:1 s.

Solution
According to Eq. (6.4.39), the state equation of the motor is
472 6 Combined Systems and System Modeling Techniques

2 3 2 3
0 1 0 0 0 
vin
x_ ¼ 40 cb =J K=J 5x þ 4 0 1=J 5 (a)
TL
0 Kb =La Ra =La 1=La 0

and the matrix output equation is

0 1 0
y¼ x (b)
0 0 K

where
0 1 0 1
x1 θ  
@ A @ _ A y θ_
x ¼ x2 ¼ θ ; y ¼ 1 ¼ (c)
y2 T
x3 ia

Based on the state and output equations, a Simulink model of the motor is generated; see
Figure 6.6.9, where a State-Space block with the name DC Motor is connected to a Mux
block, which is for inputs vin and TL , and a Demux block, which is for the outputs θ_ and T. Here,
Mux and Demux are from the Commonly Used Blocks Library and State-Space is from the
Continuous Library. Double-click the State-Space block to enter the parameters of
the block, which are the elements of the matrices in Eqs. (a) and (b); see Figure 6.6.10, where
the parameter values given in Eq. (a) in Example 6.6.2 have been used. Compared with the model
in Example 6.6.2, the state-space model here is more efficient in dealing with the multi-input–
multi-output system.
Now, run the simulation with vin ¼ 10 V, TL ¼ 0:6 N  m applied at t ¼ 0:04 s, and zero initial
disturbances. The computed motor rotation speed ω and the motor torque T are plotted against
time in Figure 6.6.11. The motor speed plot herein is the same as that in Figure 6.6.5. The motor
torque reaches a steady-state value of 0.61 N  m, which, by T ¼ Kia , is in accordance with the
current plot in Figure 6.6.5.

Figure 6.6.9 Simulink model of the DC motor with a State-Space block

 6.6 Modeling and Simulation by Simulink 473

Figure 6.6.10 Parameters of the State-Space block

For convenience in data entry and error detection, symbols can be used in specification of the
parameters for the State-Space block; for example, see Figure 6.6.12. These parameters in the
Block Parameters window, however, need to be defined in the MATLAB Command Window
before the simulation can be run. (Simulink gets the parameter values from the MATLAB
workspace in simulation.) In the current example, this is done by typing:
≫ La = 1.9e-3; Ra = 0.56; K = 0.06; Kb = K;
≫ J=9e-5; cb = 1.2e-4;
474 6 Combined Systems and System Modeling Techniques

180
160
140
120
100
ω (rad/s)

80
60
40
20
0
–20
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1

0.8
0.7
0.6
0.5
T (N.m)

0.4
0.3
0.2
0.1
0

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1

t (s)

Figure 6.6.11 Response of the motor in case 2 of Example 6.6.2

 6.7 Modeling and Simulation by Mathematica 475

Figure 6.6.12 Parameters of the State-Space block with symbols

6.7 Modeling and Simulation by Mathematica

In this section, the use of Wolfram Mathematica in modeling and simulation of dynamic
systems is illustrated in the same examples as discussed in Section 6.6. Therefore, this section
concentrates on the implementation of the mathematical models of the dynamic systems
presented in examples 6.6.1 through 6.6.4 in Wolfram Language. For an analysis of the
results, refer to the appropriate examples in Section 6.6. Refer to Appendix C on how to use
Mathematica.
Governing equations are key to modeling dynamic systems in Mathematica. The system
response can be derived by either directly using differential equation solvers (analytical
DSolve or numerical NDSolve), or by using StateSpaceModel function and correspond-
ing OutputResponse or StateResponse.
476 6 Combined Systems and System Modeling Techniques

Example 6.7.1
Consider the system in Example 6.5.3, with zero initial conditions, and inputs of the system being
u ¼ 1:5 sinð10tÞ and v ¼ 0:6:

x þ 5x þ 0:1x3  y ¼ u
4€ (a)
y_ þ 3y  6xy ¼ v (b)

Plot the system outputs x and y for 0 ≤ t ≤ 20.

Solution
Since this system of differential equations is nonlinear, we do not expect a closed-form solution;
hence, we need to use a nonlinear solver, as shown in Figure 6.7.1, obtaining the same plots (see
Figure 6.7.2) as previously shown in Figure 6.6.2 for Example 6.6.1.
Note that using a variant NDSolveValue of the Mathematica powerful numerical differential
equation solver NDSolve facilitates access to the generated responses xðtÞ and yðtÞ and their
subsequent visualization with Plot function (see Figure 6.7.2).

Figure 6.7.1 Code for Example 6.7.1

 6.7 Modeling and Simulation by Mathematica 477

0.10

0.08

0.06
x

0.04

0.02

0
0 2 4 6 8 10 12 14 16 18 20
time
0.25

0.20

0.15
y

0.10

0.05

0
0 2 4 6 8 10 12 14 16 18 20
time

Figure 6.7.2 Response (x and y) of the system in Example 6.7.1

Example 6.7.2
An armature-controlled DC motor is governed by the following differential equations as shown in
Example 6.5.4:
8
<J θ€ þ cb θ_ ¼ Kia  TL
:La dia þ Ra ia ¼ vin ðtÞ  Kb θ_
dt
where θ is the rotation angle of the motor; vin ðtÞ and ia are the voltage input and current of the
armature circuit, respectively; K is the motor torque constant; Kb is the back emf constant; and TL
denotes a load torque. As shown in Eq. (6.4.10), K ¼ Kb .
478 6 Combined Systems and System Modeling Techniques

Given the following parameter values:

La ¼ 1:9  103 H; Ra ¼ 0:56 Ω; K ¼ Kb ¼ 0:06 N  m=A

J ¼ 9  105 kg  m2 ; cb ¼ 1:2  104 N  m  s=rad

and the applied armature voltage vin ðtÞ ¼ 10uðtÞ V, plot the rotation speed ωðtÞ ¼ θðtÞ _ and the
armature current ia ðtÞ of the motor, for 0 ≤ t ≤ 0:1 s. In the simulation, the following two cases are
considered:

Case 1: the motor idles with no load torque, TL ¼ 0

Case 2: the motor idles initially, and a constant load torque is applied with a delay of 0.04 s,
TL ¼ 0:6uðt  0:04Þ Nm

Solution
For this system, a closed-form solution is expected; hence, the analytical solver DSolve
needs to be used as shown in Figure 6.7.3. Note that the solver generates functions

Figure 6.7.3 Code for Example 6.7.2 (case 1)

 6.7 Modeling and Simulation by Mathematica 479

_ needs to be plotted in addition to the armature current

θðtÞ and ia ðtÞ while the rotation speed θðtÞ
ia ðtÞ. Therefore, the derived solution θðtÞ needs to be differentiated before being used as an input to
Plot. Also, note that since two cases need to be considered, an efficient code implementation
involves setting up equations in a general form, assigning numerical values to the variables and
inputs that do not change from case 1 to case 2, but leaving the load torque as a symbolic variable.
Its value is assigned with replacement syntax when the solver for case 1 is called.
The generated responses, shown in Figure 6.7.4, are the same as previously shown in Figure
6.6.4 for Example 6.6.2.
To solve case 2, we can use the same Mathematica notebook, immediately calling the
solver (see Figure 6.7.5) since the equations and numerical values are already in memory.
(Note the replacement syntax and representation of a load torque input with a delayed
step function.)

150
ω (rad/s)

100

50

0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
time

10
ia (A)

0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
time

Figure 6.7.4 Response of the motor without a load (case 1)

The plots of θ_ ðtÞ and ia ðtÞ are shown in Figure 6.7.6. As expected, the generated responses are
the same as previously shown in Figure 6.6.5 for Example 6.6.2.
480 6 Combined Systems and System Modeling Techniques

Figure 6.7.5 Code for Example 6.7.2 (case 2)

150
ω (rad/s)

100

50

0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
time

10
ia (A)

0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
time

Figure 6.7.6 Response of the motor under a constant load torque (case 2) in Example 6.7.2
 6.7 Modeling and Simulation by Mathematica 481

Example 6.7.3
Consider a time-variant mechanical system with displacement x governed by the following
ordinary differential equation:

x þ c0 x_ þ ðk0  ε sin ΩtÞx ¼ q

m€ (a)

and the system parameter values as follows:

m ¼ 2; c0 ¼ 3; c1 ¼ 0:1; k0 ¼ 32; ε ¼ 12; and Ω ¼ 10:

The external force is given as q ¼ 2 sinð4tÞ. Plot the displacement of this system for the time
period of 0 ≤ t ≤ 15.
Solution
This system can be solved using the numerical differential equation solver NDSolve (for a time-
variant ordinary differential equation, no closed-form solution is expected) in the same manner as
for Example 6.7.2. This code is shown in Figure 6.7.7.
A more efficient implementation uses the Modelica System Model functionality (not available
for Mathematica versions older than 12), as shown in Figure 6.7.8. Since this code is in the new
notebook, setting up the symbolic equations and assigning numerical values to the variables is
necessary.

Figure 6.7.7 Code for Example 6.7.3: using a numerical differential equation solver
482 6 Combined Systems and System Modeling Techniques

Figure 6.7.8 Code for Example 6.7.3: using the Modelica System Model

0.2

0.15

0.1

0.05
x

–0.05

–0.1

–0.15

–0.2
0 5 10 15
time

Figure 6.7.9 Displacement of the time-variant system in Example 6.7.3

The same graph is generated by both methods. This plot (see Figure 6.7.9) is identical to the one
produced for Example 6.6.3 and shown in Figure 6.6.8.
 6.7 Modeling and Simulation by Mathematica 483

Example 6.7.4
This example shows a computation of the system response deriving its state-space representation
first.
The Mathematica family of functions that generate state-space models is very powerful and can
handle a wide range of dynamic systems: linear and nonlinear, time-invariant and time-varying,
continuous and discrete. Any of these models is then used as an argument for the response-
generating functions: OutputResponse (when only a specific output is of interest) and
StateResponse (when all state variables need to be monitored and analyzed).
Since the armature-controlled DC motor of Example 6.7.2 is a linear dynamic system, it can be
modeled using StateSpaceModel and the corresponding OutputResponse functions.

(a)

A B

0· 1 0· 0· 0· S
C K 1
0· – —b — 0· –—
J J J
Out[•] = Kb Ra 1
0· – — – — — 0·
La La La
0· 1 0· 0· 0·
0· 0· K 0· 0·

C D
(b)

Figure 6.7.10 Solving Example 6.7.4 with StateSpaceModel function: (a) generating the
state-space representation from governing equations; (b) visualizing state-space matrices
484 6 Combined Systems and System Modeling Techniques

Nonetheless, if the execution time is crucial, an analytical differential equation solver may yield a
better performance.
Unlike with MATLAB/Simulink, the derivation of a state-space model by hand is not necessary – it
can be done automatically with the software, as shown in Figure 6.7.10.
The derivation of a state-space model in symbolic form is recommended so that its form can be
analyzed, and its correctness verified (see Eqs. (a), (b), and (c) of Example 6.6.4).
To generate and plot the required response, symbolic variables must have assigned values
before OutputResponse is called. This code is shown in Figure 6.7.11 (it follows the code in
Figure 6.7.10 in the same notebook).
The generated plots (see Figure 6.7.12) are identical to those produced for Example 6.6.4 and
shown in Figure 6.6.11.

Figure 6.7.11 Code for Example 6.7.4: generating the required outputs from the derived
state-space model and plotting them
 Chapter Summary 485

Figure 6.7.12 Response of the motor in case 2 of Example 6.7.2

CHAPTER SUMMARY
This chapter discusses modeling of a dynamic system consisting of multiple subsystems,
typically of a different nature. Typical system modeling approaches such as transfer func-
tions, state-space representations, and block diagrams in the s-domain and the time domain
are discussed. Foundations of electromechanical systems are presented on the example of a
DC motor. The derived mathematical model is implemented in software: MATLAB/
Simulink and Wolfram Mathematica; the system behavior in response to several standard
inputs is simulated and analyzed.
Upon completion of this chapter the reader should be able to:

(1) Understand the four types of a dynamic system model: governing equations, state-space
representations, transfer functions, and block diagrams, and derive any three of them
from the given one.
(2) Construct a block diagram of a dynamic system from the given governing equa-
tions or state-space model, expanding it to the degree necessary for a good under-
standing of the input signal propagation through the system and its transformation
into system output(s).
486 6 Combined Systems and System Modeling Techniques

(3) Recollect their knowledge of electromagnetism and fundamental physical laws govern-
ing D’Arsonval motion and, subsequently, the operation of a DC motor.
(4) Understand the interaction between electrical and mechanical subsystems of an electro-
mechanical system, derive models of armature-controlled and field-controlled DC
motors, and of a DC motor with a load.
(5) Implement the derived mathematical models in MATLAB/Simulink and Wolfram
Mathematica software, simulate the system behavior in response to standard inputs,
and analyze the generated results.

REFERENCES
1. J. L. Meriam, L. G. Kraige, and J. N. Bolton, Engineering Mechanics: Dynamics, 8th ed., Wiley,
2015.
2. N. Lobontiu, System Dynamics for Engineering Students: Concepts and Applications, 2nd ed.,
Academic Press, 2017.
3. W. J. Palm, III, System Dynamics, 3rd ed. McGraw-Hill, 2014.
4. C. M. Close, D. K. Frederick, and J. C. Newell, Modeling and Analysis of Dynamic Systems, Wiley,
3rd ed., 2002.
5. D. G. Luenberger, Introduction to Dynamic Systems: Theory, Models, and Applications, Wiley,
1979.
6. R. S. Esfandiari and B. Lu, Modeling and Analysis of Dynamic Systems, CRC Press, 2010.
7. R. L. Woods and K. L. Lawrence, Modeling and Simulation of Dynamic Systems, Prentice Hall,
1997.
8. K. Ogata, System Dynamics, 4th ed., Pearson, 2004.

PROBLEMS
Section 6.2 System Modeling Techniques
6.1 Derive the transfer function for the following dynamic systems, where f ðtÞ is the input
and xðtÞ is the output. Represent these transfer functions in zero-pole-gain form.
Indicate the order of each transfer function. Assume that f ð0 Þ ¼ f_ ð0 Þ ¼ f€ð0 Þ ¼ 0.
x þ 11x_ þ 6x ¼ 3f ðtÞ þ f_ ðtÞ
(a) xð3Þ þ 6€
ð3Þ
(b) x þ 13€ x þ 50x_ þ 56x ¼ f€ðtÞ  f ðtÞ
(c) 6x þ 10xð3Þ þ 15€
ð4Þ
x þ 22x_ þ 15x ¼ 12f ðtÞ
x þ 7x_ þ 5x ¼ 15f ðtÞ þ 3f_ ðtÞ
(d) 3xð3Þ þ 6€
x þ 7x_ þ 3x ¼ 9f ðtÞ þ 3f_ ðtÞ þ 2f€ðtÞ
(e) 6xð3Þ þ 2€
(f) x þ 7x þ 20xð3Þ þ 30€
ð5Þ ð4Þ
x þ 24x_ þ 8x ¼ 3f ðtÞ  5f_ ðtÞ þ f€ðtÞ þ f ð3Þ ðtÞ
6.2 Derive the transfer function for the following dynamic systems, where f ðtÞ is the input
_ is the output. Indicate the order of each transfer function. Assume that
and 2xðtÞ þ 3xðtÞ

f ð0 Þ ¼ 0.
(a) x€ þ 3x_ þ 7x ¼ 5f ðtÞ
(b) 12xð3Þ þ 5€x þ 2x_ þ 6x ¼ 24f_ ðtÞ  18f ðtÞ
 Problems 487

x þ 5x_ þ 3x ¼ 18f ðtÞ þ 3f_ ðtÞ

(c) 6xð4Þ þ 6xð3Þ þ 9€
x þ 9x_ þ 8x ¼ 8f ðtÞ þ 10f_ ðtÞ
(d) 2€

6.3 Derive the following transfer functions:

x€ þ 3x_ þ 7x  7y ¼ 5f ðtÞ
(a)
2€y þ 7y  7x ¼ 0
where the input is f ðtÞ, the outputs are xðtÞ and yðtÞ
5€x þ 2x_  2y_ þ 11x  8y ¼ f ðtÞ
(b)
3€y þ 2y_ þ 8y  2x_  8x ¼ 0
_
where the input is f ðtÞ, the outputs are xðtÞ  yðtÞ and yðtÞ
2€x þ x_  y_ þ 4x ¼ f ðtÞ
(c)
y€ þ 3y_ þ y  x_ ¼ gðtÞ
_
where the inputs are f ðtÞ and gðtÞ, the output is yðtÞ
2€x þ 3x_  y_ þ 4x  3y ¼ f ðtÞ
(d)
6€y þ 2y_ þ 5y  x_ ¼ gðtÞ
_ and yðtÞ
where the inputs are f ðtÞ and gðtÞ, the outputs are xðtÞ _
6.4 Derive the state-space representation for the following dynamic systems, where f ðtÞ is
the input and xðtÞ is the output. Also, cast the state and output equations in matrix form.
(a) xð3Þ þ 6€x þ 11x_ þ 6x ¼ 3f ðtÞ
ð3Þ
(b) x þ 13€ x þ 50x_ þ 56x ¼ 15f ðtÞ
(c) 6xð4Þ þ 10xð3Þ þ 15€ x þ 22x_ þ 15x ¼ 12f ðtÞ
(d) 3xð3Þ þ 6€ x þ 7x_ þ 5x ¼ 15f ðtÞ þ 3f_ ðtÞ
6.5 Derive the state-space representation for the following dynamic systems, where f ðtÞ is
the input and 5xðtÞ þ xðtÞ _ is the output. Also, cast the state and output equations in
matrix form.
(a) x€ þ 3x_ þ 7x ¼ 5f ðtÞ
(b) 12xð3Þ þ 5€ x þ 2x_ þ 6x ¼ 24f ðtÞ
(c) 6xð4Þ þ 6xð3Þ þ 9€ x þ 5x_ þ 3x ¼ 18f ðtÞ
6.6 Derive the state-space representations for the systems in Problem 6.3; cast the state and
output equations in matrix form.
6.7 Derive the state-space representations for the following systems described by their
transfer functions:
YðsÞ 6
(a) ¼ 2
RðsÞ s þ 4s þ 17
YðsÞ 15
(b) ¼
RðsÞ 2s3 þ 5s2 þ 3s þ 4
YðsÞ 7
(c) ¼ 4
RðsÞ 6s þ x þ 9s2 þ 2s þ 1
3
488 6 Combined Systems and System Modeling Techniques

6.8 Construct a block diagram for the dynamic system represented by the following govern-
ing equations:
x_ þ 3x  5y ¼ 2f ðtÞ
(a)
y_ þ 3y þ 4x ¼ 0
where the input is f ðtÞ, the output is y
x_ þ 2x  6y ¼ 3f ðtÞ
(b)
y_ þ 8y  4x ¼ 5f ðtÞ
where the input is f ðtÞ, the output is y
8
>x_  2y ¼ 0
>
<
y_ þ 7x þ 4y  3w ¼ 0
(c)
>
> w_  5v ¼ 0
:
v_  6x þ 2w ¼ f ðtÞ
where the input is f ðtÞ, the output is x
x þ 6x_ þ 4x ¼ 3f ðtÞ þ 5gðtÞ
(d) 2€
where the inputs are f ðtÞ and gðtÞ, the output is x
(e) where the inputs are f ðtÞ and gðtÞ, the output is y
x€ þ 5x_ þ 7y ¼ f ðtÞ þ 6gðtÞ
(f)
y_ þ 2x  3y ¼ 4gðtÞ
where the inputs are f ðtÞ and gðtÞ, the output is x
X ðsÞ
6.9 Derive the transfer function for the dynamic systems modeled with the block
RðsÞ
diagrams shown in Figure P6.9:
 Problems 489

R(s) + + X(s)
G1 G2
− −

(a)

R(s) + + X(s)
G1 G2
− −

H2

H1

(b)

G2

+
R(s) + + Y(s) X(s)
G1 G3 G4
− −

(c)
G2
+
R(s) + + Y(s) X(s)
G1 G3 G4
− −

H2
H1

(d)

H2
R(s) + + + X(s)
G1 G2
− − −

H1

H3

(e)

Figure P6.9
490 6 Combined Systems and System Modeling Techniques

X ðsÞ X ðsÞ
6.10 Derive the transfer functions and for the dynamic systems modeled with the
RðsÞ DðsÞ
block diagrams shown in Figure P6.10:

H1
D(s)
− −
R(s) + + X(s)
G1 G2
−

H2

(a)

H1
D(s)
− −
R(s) + + X(s)
G1 G2 G3
−

H2

(b)

D(s)
−
R(s) + + + X(s)
G1 G2 G3
− −

H2

H1

(c)

R(s) + X(s)
G1 G2
−

+
−
H1 H2
+
D(s)

(d)

Figure P6.10
 Problems 491

Section 6.3 Fundamentals of Electromechanical Systems

6.11 A transducer – an electromechanical system designed to measure translational motion –
is shown in Figure P6.11. The permanent magnet produces a uniform magnetic field with
magnetic flux density ~B. This magnet is in one-dimensional motion, with displacement
_
xðtÞ as shown and corresponding velocity xðtÞ. A conductor is fixed in space in such a
way that its portion of the length h is inside the air gap between the north and south
poles of the permanent magnet. This conductor is modeled as a series connection of an
inductor with inductance L0 and resistor with resistance R0 . Another resistor with
resistance R is connected to this conductor with a lossless wire, forming a closed circuit
as shown in Figure P6.11.

L0 R0
x(t) N
+
B h εb R
− i
S
R
i
(a) (b)

Figure P6.11 Transducer: (a) schematic, (b) electrical subsystem

When the permanent magnet experiences translational motion, the magnetic environ-
ment of the conductor changes, hence, an emf is induced in this conductor according to
Faraday’s law of induction, and current starts flowing in the circuit.
(a) Derive the governing equations for this system.
VR ðsÞ
(b) Derive the transfer function , considering that the input to the system is
X ðsÞ
the displacement of the magnet xðtÞ and the output is the voltage drop vR on the
resistor R.
6.12 Consider the transducer described in Problem 6.11.
The permanent magnet of mass M is sliding forward along a stationary horizontal
surface under the influence of the applied force fM ðtÞ. The direction of motion and the
displacement of the magnet xðtÞ are shown in Figure P6.12. There is an oil film with a
viscous friction coefficient c between the magnet and the surface.
(a) Derive the governing equations for this system.
VR ðsÞ
(b) Derive the transfer function , considering that the input to the system is the
FM ðsÞ
applied force fM ðtÞ and the output is the voltage drop vR on the resistor R.
492 6 Combined Systems and System Modeling Techniques

fM (t)
M
x(t) N

B h

S
c

R
i

Figure 6.12 Transducer for Problem P6.12

6.13 Consider the system shown in Figure P6.13. A conductor of length h is attached to a
block of mass M with a rigid dielectric connector. The block is sliding along a station-
ary horizontal surface in the direction shown by its displacement xðtÞ. The block is
constrained with the linear spring with stiffness coefficient k, and is separated from the
surface by an oil film with viscous friction coefficient c.

L0 R0 N

B h
+ +
e in εb S
i x(t)
− −
M + −
e in
i
c
k
(a) (b)

Figure P6.13 Conductor in the magnetic field: (a) electrical subsystem, (b) schematic

The conductor is fixed in space in such a way that its length h is inside the air gap
between the north and south poles of the permanent magnet, which produces a uniform
magnetic field with magnetic flux density ~ B. The conductor is modeled as a series
connection of an inductor with inductance L0 and resistor with resistance R0 . The
conductor is connected to a voltage source, which provides the input voltage ein ðtÞ.
(a) Derive the governing equations for this system.
(b) Derive the state-space representation, considering that the system input is voltage
ein ðtÞ, and the output is the velocity of the block.
6.14 A setup for a magnetic levitation laboratory experiment (Yeh et al., Sliding Control of
Magnetic Bearing Systems, ASME Journal of Dynamic Systems, Measurement, and
Control, Vol. 123, September 2001) is shown in Figure P6.14.
 Problems 493

The current-carrying coil, wound around a metal core, constitutes an electromagnet

that exerts an attracting electromagnetic force Fem on the metal ball of mass m, making
it levitate.
Fem
Metal core L0 R0

+
Coil + x(t)
e in
e in m
− i
−

h
x(t) mg

m Levitating
metal ball

(a) (b) (c)

Figure P6.14 Magnetic levitation experiment: (a) schematic, (b) electrical subsystem, (c) mechanical subsystem

The electromagnetic force is modeled as follows:

i2
Fem ¼ KF
ðh  xÞ2
where KF is a constant that depends on the construction of the electromagnet (material
of the core and coil, number and geometry of loops in the coil, etc.), xðtÞ is the position
of the levitating ball, measured upward from a fixed at-rest position, and h is a constant
air-gap distance from the electromagnet core to the ball.
(a) Derive the governing equations for this system.
(b) Derive the state-space representation, considering that the system input is voltage
ein ðtÞ, and the output is the displacement of the ball.
6.15 A solenoid actuator, typically used in hydraulic or pneumatic systems, is an electro-
mechanical device that converts electrical energy into a translational force for pushing/
pulling a valve.
The schematic of a typical solenoid actuator (Chladny et al., Modeling Automotive
Gas-Exchange Solenoid Valve Actuators, IEEE Transactions on Magnetics, Vol. 41,
March 2005) is shown in Figure P6.15.
The push-type solenoid actuator, shown in Figure P6.15(a), consists of a wire coil,
wound around an iron core – a plunger, which is connected to a valve of mass m with a
rigid push-pin. Applying voltage ein ðtÞ to the coil makes it into an electromagnet, which
exerts a force Fem on the plunger, pushing it to the right toward the center of the coil
(the displacement of the plunger-valve assembly is xðtÞ as shown in Figure P6.15(c)). A
spring with stiffness coefficient k returns the valve to its original position when the
current i no longer flows. A viscous friction force acts on the moving valve; the viscous
friction coefficient is c (as shown in Figure P6.15(c)).
494 6 Combined Systems and System Modeling Techniques

Figure P6.15 Solenoid actuator: (a) schematic, (b) electrical subsystem, (c) mechanical subsystem

The coil-plunger assembly is modeled as a series connection of the resistor R and the
inductor. The inductance of the solenoid actuator coil is modeled as a nonlinear
function of the plunger displacement: it increases when the plunger moves toward the
center of the coil and decreases as the plunger moves out of the coil. This inductance is
represented as

L0
LðxÞ ¼
1  x=d

where L0 and d are constants that depend on the geometry and material of the coil.
Since the solenoid coil is considered to be an ideal inductor, a linear relationship
exists between the current through the inductor iL and magnetic flux ϕ: ϕ ¼ LiL (see Eq.
(4.1.18)). The time derivative of the magnetic flux is equal to the voltage drop across the
dϕ
inductor: ¼ vL (see Eq.( 4.1.20)).
dt
The electromagnetic force Fem is modeled as the first derivative of the energy stored in
dW
an inductor with respect to displacement xðtÞ: Fem ¼ , where W is described by Eq.
dx
(4.1.22).
Assume that the return spring is at its free length and the electromagnetic force is zero
when x ¼ 0.
(a) Derive the governing equations for this system.
(b) Derive the state-space representation, considering that the system input is the
voltage ein ðtÞ, and the output is the displacement of the valve.
 Problems 495

6.16 Consider the solenoid actuator described in Problem 6.15.

Experimentation led to development of the following expressions for the solenoid
current and the electromagnetic force:

iL ðϕÞ ¼ a1 ϕ3 þ a2 ϕ
Fem ðϕÞ ¼ b1 ϕ6 þ b2 ϕ4 þ b3 ϕ2

where a1 ; a2 ; b1 ; b2 ; b3 are empirical constants (Vaughan, Gamble, The Modeling

and Simulation of a Proportional Solenoid Valve, ASME Journal of Dynamic Systems,
Measurement, and Control, Vol. 118, March 1996).
Assume that the return spring is at its free length and the electromagnetic force is
zero when x ¼ 0.
(a) Derive the governing equations for this system.
(b) Derive the state-space representation, considering that the system input is the
voltage ein ðtÞ, and the output is the displacement of the valve.

Section 6.4 Direct Current Motors

6.17 An armature-controlled DC motor is used to rotate the load JL as shown in Figure
P6.17. The load is connected to the motor with a very stiff massless shaft and is
supported by bearings with a damping coefficient bL .
The system input is the voltage ein ðtÞ supplied to the armature circuit, while the
output is the rotational velocity of the load ωL .

Armature circuit
bm bL
+ La
Ra +
ein ia εb Jm JL
− −
Tm ωm ωL
Electrical subsystem Mechanical subsystem

Figure P6.17 Armature-controlled DC motor with a load

(a) Derive the governing equations for this system.

ΩL ðsÞ
(b) Derive the transfer function .
Ein ðsÞ
6.18 An armature-controlled DC motor is used to rotate the load JL as shown in Figure
P6.18. The load is connected to the motor through a converter Ka and is supported by
bearings with a damping coefficient bL . The converter transforms the rotational vel-
ocity ωm of the motor into torque TL applied to the rotor in the following manner:
TL ¼ Ka ωm , where Ka is a known converter constant. The converter and the connector
shafts are considered massless.
496 6 Combined Systems and System Modeling Techniques

Armature circuit
bm bL

+ La
Ra +
ein ia εb Jm Ka JL
− −
Tm ωm TL ωL
Electrical subsystem Mechanical subsystem

Figure P6.18 Armature-controlled DC motor with a load and converter

The system input is the voltage ein ðtÞ supplied to the armature circuit, while the
output is the rotational velocity of the load ωL .
(a) Derive the governing equations for this system.
ΩL ðsÞ
(b) Derive the transfer function .
Ein ðsÞ
(c) Construct a modular block diagram and indicate all the relevant signals such as the
armature current ia , back emf εb , motor torque Tm , motor velocity ωm , bearing
torques Tbm and TbL , load torque TL , and load velocity ωL .
6.19 An armature-controlled DC motor is used to rotate the load JL as shown in Figure
P6.19. The load is supported by bearings with a damping coefficient bL , and is con-
nected to the motor with a flexible shaft, modeled as a torsional spring with a stiffness
coefficient kT .
The system input is the voltage ein ðtÞ supplied to the armature circuit, while output is
the rotational velocity of the load ωL .
Armature circuit
bm bL

+ La KT
Ra +
ein ia εb Jm JL
− −
Tm ωm ωL
Electrical subsystem Mechanical subsystem

Figure P6.19 Armature-controlled DC motor with a load and a flexible shaft

(a) Derive the governing equations for this system.

ΩL ðsÞ
(b) Derive the transfer function .
Ein ðsÞ
(c) Derive the state-space representation.
(d) Construct a modular block diagram and indicate all the relevant signals such as the
armature current ia , back emf εb , motor torque Tm , motor velocity ωm , bearing
torques Tbm and TbL , spring torque Tk , and load velocity ωL .
 Problems 497

6.20 A field-controlled DC motor is used to rotate the load JL as shown in Figure P6.20. The
load is supported by bearings with a damping coefficient bL , and is connected to the
n2
motor through a gear pair with a gear ratio N ¼ . The shafts connecting the gear pair
n1
with the motor and the load are short, stiff, and considered massless.
The system input is the voltage ein ðtÞ supplied to the field circuit, while output is the
rotational velocity of the load ωL .
Field circuit
bm
+ Lf n1
ein bL
Rf Jm
− if

Tm θm JL
Electrical subsystem
n2
θL ωL
Mechanical subsystem

Figure P6.20 Field-controlled DC motor with a load and gear pair

(a) Derive the governing equations for this system.

ΩL ðsÞ
(b) Derive the transfer function .
Ein ðsÞ
(c) Derive the state-space representation.
6.21 A field-controlled DC motor is used to rotate the load JL as shown in Figure P6.21. The
load is supported by bearings with a damping coefficient bL , and is connected to the
motor with a flexible shaft, modeled as a torsional spring with a stiffness coefficient kT .
The load is subject to an unknown drag torque TD .
The system inputs are the voltage ein ðtÞ supplied to the field circuit and the drag
torque TD , while output is the rotational velocity of the load ωL .

Figure P6.21 Field-controlled DC motor with a load, flexible shaft, and drag torque
498 6 Combined Systems and System Modeling Techniques

(a) Derive the governing equations for this system.

ΩL ðsÞ ΩL ðsÞ
(b) Derive the transfer functions and .
Ein ðsÞ TD ðsÞ
(c) Derive the state-space representation.
(d) Construct a modular block diagram and indicate all the relevant signals such as the
field current if , motor torque Tm , motor displacement θm , bearing torques Tbm and
TbL , spring torque Tk , drag torque TD , load velocity ωL , and displacement θL .
6.22 A field-controlled DC motor is used to rotate the load JL as shown in Figure P6.22. The load
is connected to the motor through a converter Ka and is supported by bearings with a
damping coefficient bL . The converter transforms the rotational velocity ωm of the motor
into the torque TL applied to the rotor in the following manner: TL ¼ Ka ωm , where Ka is a
known converter constant. The converter and the connector shafts are considered massless.
The load is subject to an unknown drag torque TD .
The voltage supplied to the field circuit of the DC motor is driven by a generator.
This voltage is proportional to the current in the generator circuit as follows: ef ¼ Kg ig ,
where a Kg is a known constant.
The system inputs are the voltage ein ðtÞ supplied to the generator and the drag torque
TD , while output is the rotational velocity of the load ωL .

Generator circuit Field circuit

+ Lg + Lf
ein Rg ef Rf
− ig − if

Electrical subsystem 2 Electrical subsystem 1

bm bL TD

Jm Ka JL

Tm ωm TL ωL
Mechanical subsystem

Figure P6.22 Field-controlled DC motor with a load and a generator

(a) Derive the governing equations for this system.

ΩL ðsÞ ΩL ðsÞ
(b) Derive the transfer functions and .
Ein ðsÞ TD ðsÞ
(c) Derive the state-space representation.
(d) Construct a modular block diagram and indicate all the relevant signals such as
the generator current ig , field circuit input voltage ef , field current if , motor torque
Tm , motor velocity ωm , bearing torques Tbm and TbL , drag torque TD , and load
velocity ωL .
 Problems 499

Section 6.5 Block Diagrams in the Time Domain

6.23 Construct time-domain block diagrams for the dynamic systems represented by equa-
tions in Problem 6.8.
6.24 Construct a time-domain block diagram for the DC-motor system in Problem 6.18.
6.25 Construct a time-domain block diagram for the DC-motor system in Problem 6.19.
6.26 Construct a time-domain block diagram for the DC-motor system in Problem 6.20.
6.27 Construct a time-domain block diagram for the DC-motor system in Problem 6.21.

Section 6.6 Modeling and Simulation by Simulink

6.28 Consider that the transducer described in Problem 6.11 has the following parameters:

L0 ¼ 0:003 H; R0 ¼ 0:8 Ω; R ¼ 1:2 Ω; h ¼ 0:03 m; B ¼ 1:25 T:

The input displacement of the magnet is xðtÞ ¼ t.
Plot the voltage drop on the resistor R (consider a simulation time 0 ≤ t ≤ 0:02 s), find
the steady-state value if it exists, and compute the settling time for 0.5% allowed error.
6.29 Consider that the transducer described in Problem 6.12 has the following parameters:

L0 ¼ 0:003 H; R0 ¼ 0:8 Ω; R ¼ 1:2 Ω; h ¼ 0:03 m; B ¼ 1:25 T;

N
m ¼ 0:25 kg; c ¼ 7:5  104
Am
The external force applied to the magnet is constant: fM ðtÞ ¼ 3:1 N.
Plot the voltage drop on the resistor R (consider a simulation time 0 ≤ t ≤ 200s), find the
steady-state value if it exists, and compute the settling time for 1.0% allowed error.
6.30 Consider that the armature-controlled DC motor with a load and converter system
described in Problem 6.18 has the following parameters:
ðNmsÞ
La ¼ 0:004H; Ra ¼ 0:75 Ω; Jm ¼ 8:0  103 kgm2 ; bm ¼ 0:005 ;
rad
Nm Vs ðNmsÞ
Km ¼ 0:3 ; Kb ¼ 0:2 ; Ka ¼ 0:4 ;
A rad rad
Nms
JL ¼ 12:0  103 kgm2 ; bL ¼ 0:006
rad

(a) The voltage applied to the armature circuit of the DC motor is constant:
ein ðtÞ ¼ 0:1 V. Plot the velocity of the load ωL (consider a simulation time
0 ≤ t ≤ 20 s), and find the steady-state value if it exists.
(b) The voltage applied to the armature circuit of the DC motor is a positive half-sine
periodic function ein ðtÞ ¼ 0:5jsinðtÞj V (jsinðtÞj indicates the absolute value of the sine
function). Plot the velocity of the load ωL (consider simulation time 0 ≤ t ≤ 50 s),
and find the largest velocity value after the amplitude of the oscillations settles to a
constant sinusoidal pattern.
500 6 Combined Systems and System Modeling Techniques

31 Consider that the field-controlled DC motor with a load and gear pair described in
Problem 6.20 has the following parameters:
ðNmsÞ
Lf ¼ 0:006 H; Rf ¼ 0:56 Ω; Jm ¼ 9:1  104 kgm2 ; bm ¼ 0:0012 ;
rad
Nm n2 ðNmsÞ
Km ¼ 0:06 ; N ¼ ¼ 2:5 ; JL ¼ 15:0  104 kgm2 ; bL ¼ 0:002
A n1 rad
(a) The voltage applied to the field circuit of the DC motor is constant: ein ðtÞ ¼ 5:2 V.
Plot the velocity of the load ωL (consider a simulation time 0 ≤ t ≤ 20 s), and find
steady-state value if exists; plot the field current (consider simulation time
0 ≤ t ≤ 0:2 s), and find the steady-state value if it exists.
(b) The voltage applied to the field circuit of the DC motor is a pulse:
ein ðtÞ ¼ 2uðtÞ  2uðt  5Þ V. Plot the velocity of the load ωL (consider a simulation
time 0 ≤ t ≤ 20 s), find the maximum velocity and its steady-state value if it exists; plot
the field current (consider a simulation time 0 ≤ t ≤ 20 s), find the maximum current
and its steady-state value if exists.
(c) The voltage applied to the field circuit of the DC motor is a trapezoidal pulse, shown
in Figure P6.31.

Figure P6.31 Input voltage

Plot the velocity of the load ωL (consider a simulation time 0 ≤ t ≤ 30 s), find the
maximum velocity and its steady-state value if it exists; plot the field current (consider
simulation time 0 ≤ t ≤ 30 s), find the maximum current and its steady-state value if it
exists.
 Problems 501

6.32 Consider that the field-controlled DC motor with a load, flexible shaft, and drag
described in Problem 6.21 has the following parameters:
ðNmsÞ
Lf ¼ 0:006 H; Rf ¼ 0:56 Ω; Jm ¼ 9:1  104 kgm2 ; bm ¼ 0:0012 ;
rad
Nm Nm ðNmsÞ
Km ¼ 0:06 ; kT ¼ 0:5 ; JL ¼ 25:0  104 kgm2 ; bL ¼ 0:002 ;
A rad rad
ein ðtÞ ¼ 5 V

(a) Plot the velocity of the load ωL (consider simulation time 0 ≤ t ≤ 15 s) for the case
when the drag torque TD ¼ 0 and when TD ¼ 0:005 θ_ L sgnðθ_ L Þ.
8
<1 when x > 0
sgnðθ_ L Þ is the signum or sign function, defined as sgnðx ¼ 1 when x < 0 .
:
0 when x ¼ 0
Do both systems converge to a steady state? Qualitatively describe the influence of
the drag torque on the system response.
(b) Plot the velocity of the load ωL (consider simulation time 0 ≤ t ≤ 15 s) for the case
when drag torque is applied with 5-s delay: TD ¼ 0:005 θ_ L sgnðθ_ L Þuðt  5Þ. Does the
system converge to a steady state? Qualitatively describe influence of the delayed
drag torque on system response.

Section 6.7 Modeling and Simulation by Mathematica

6.33 Solve Problem 6.28 by Mathematica.
6.34 Solve Problem 6.29 by Mathematica.
6.35 Solve Problem 6.30 by Mathematica.
6.36 Solve Problem 6.31 by Mathematica.
6.37 Solve Problem 6.32 by Mathematica.
7 System Response Analysis

Contents
7.1 System Response Analysis in the Time Domain 502
7.2 Stability Analysis 544
7.3 System Response Analysis in the Frequency Domain 558
7.4 Time Response of Linear Time-Varying and Nonlinear Systems 571
Chapter Summary 577
References 578
Problems 578

In the previous chapters, mathematical models of various dynamic systems have been
developed; several modeling tools, including differential equations by the first laws of nature,
transfer function formulations, block diagrams, and state-space representations, have been
introduced. With these models and tools, the dynamic response of a system can be deter-
mined, and the understanding of its behaviors can be gained through analysis and
simulation.
In this chapter, three main issues regarding physical behaviors of dynamic systems are
addressed: system response analysis in the time domain, stability analysis, and system
response analysis in the frequency domain. In the presentation, analytical methods are
used for simple systems; MATLAB and Simulink are used for general linear and nonlinear
systems.

7.1 System Response Analysis in the Time Domain

In this section, solutions of the response of linear time-invariant systems in the time domain
are determined by Laplace transform. Refer to Section 2.6 for the Laplace transform
method.

502
 7.1 System Response Analysis in the Time Domain 503

7.1.1 General Concepts

Consider a general dynamic system governed by an n-th-order linear ordinary differential
equation

AðDÞyðtÞ ¼ BðDÞrðtÞ (7.1.1)

where D ¼ d=dt, AðDÞ and BðDÞ are polynomials of operator D

X
n X
m
AðDÞ ¼ ak Dk ; BðDÞ ¼ bl Dl
k¼0 l¼0

with ak and bl being constants, and rðtÞ and yðtÞ are the forcing function (system input) and
response (system output), respectively. Refer to Section 2.5.1 for details on the operator
polynomials. Let the initial conditions of Eq. (7.1.1) be

yð0Þ ¼ a0;0 ; Dyð0Þ ¼ a0;1 ; . . . ; Dn1 yð0Þ ¼ a0;n1 (7.1.2)

where a0;0 ; a0;1 ; . . . ; a0;n1 are prescribed values. The basic problem of system response
analysis in the time domain is to find the solution of the differential equation (7.1.1) subject
to the initial conditions (7.1.2).

Free Response and Forced Response

The solution of the previous differential equation can be obtained by Laplace transforms. To
this end, performing Laplace transform of Eq. (7.1.1) yields

AðsÞYðsÞ ¼ IðsÞ þ BðsÞRðsÞ (7.1.3)

where RðsÞ and YðsÞ are the Laplace transforms of the input rðtÞ and output yðtÞ, respectively,
and IðsÞ is a polynomial of s, which is a linear combination of the initial values
a0;0 ; a0;1 ; . . . ; a0;n1 . It can be shown that IðsÞ ¼ 0 if all the initial values of system response
are set to zero. Following Section 2.6, the time response of the system is given by

yðtÞ ¼ yI ðtÞ þ yF ðtÞ (7.1.4)

with

1 IðsÞ
yI ðtÞ ¼ L (7.1.5a)
AðsÞ

yF ðtÞ ¼ L1 fGðsÞRðsÞg (7.1.5b)

where L1 is the inverse Laplace transform operator, yI ðtÞ and yF ðtÞ are the free response and
forced response, respectively, and G(s) is a transfer function of the system given by
504 7 System Response Analysis

BðsÞ bm sm þ bm1 sm1 þ . . . þ b1 s þ b0

GðsÞ ¼ ¼ (7.1.6)
AðsÞ an sn þ an1 sn1 þ . . . þ a1 s þ a0

For inverse Laplace transforms, refer to Section 2.4.4.

Example 7.1.1
A spring–mass–damper system is governed by

_ þ kxðtÞ ¼ f ðtÞ
x þ cxðtÞ
m€
_
xð0Þ ¼ x0 ; xð0Þ ¼ v0

Laplace transform of the previous differential equation yields

 2 
ms þ cs þ k X ðsÞ ¼ FðsÞ þ mðsx0 þ v0 Þ þ cx0
|fflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflffl} |ffl{zffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
AðsÞ BðsÞ¼1 IðsÞ

By Eq. (7.1.5a,b), the free response and forced response of the system are given by

mðsx0 þ v0 Þ þ cx0
yI ðtÞ ¼ L1
ms2 þ cs þ k

1 1
yF ðtÞ ¼ L FðsÞ
ms2 þ cs þ k

From a physical viewpoint, yI ðtÞ is the system response excited by initial disturbances and
it is described by the differential equation

AðDÞyI ðtÞ ¼ 0 (7.1.7a)

subject to initial conditions

yI ð0Þ ¼ a0;0 ; DyI ð0Þ ¼ a0;1 ; . . . ; Dn1 yI ð0Þ ¼ a0;n1 (7.1.7b)

The yF ðtÞ, on the other hand, is the system response excited by the input rðtÞ and it is
governed by

AðDÞyF ðtÞ ¼ BðDÞrðtÞ (7.1.8a)

with zero initial conditions

yF ð0Þ ¼ 0; DyF ð0Þ ¼ 0; . . . ; Dn1 yF ð0Þ ¼ 0 (7.1.8b)

Because Eq. (7.1.7a) is a homogeneous differential equation, its solution can be expressed by

yI ðtÞ ¼ A1 eλ1 t þ A2 eλ2 t þ . . . þ An eλn t (7.1.9)

 7.1 System Response Analysis in the Time Domain 505

where λk are the roots of the characteristic equation

AðλÞ ¼ an λn þ an1 λn1 þ . . . þ a1 λ þ a0 ¼ 0

and Ak are constants that are eventually determined by the initial conditions (7.1.7b). Here,
distinct characteristic roots have been assumed. If AðλÞ has l repeated roots λ0 (a root of
l multiplicity) and m distinct roots λ1 ; λ2 ; . . . ; λm , with l þ m ¼ n and λ0 ≠ λk for
j ¼ 1; 2; . . . ; m, the free response is of the form
 
yI ðtÞ ¼ A1 eλ1 t þ A2 eλ2 t þ . . . þ Am eλm t þ B1 tl1 þ B2 tl2 þ . . . þ Bl eλ0 t (7.1.10)

Refer to Section 2.6 for more on the complementary homogeneous equation.

The solution of Eq. (7.1.8a), through the application the convolution theorem
(Section 2.4.3) to Eq. (7.1.5b), can be expressed by the convolution integral
ðt
yF ðtÞ ¼ gðt  τÞrðτÞdτ (7.1.11)
0

Here gðtÞ is the impulse response function of the system, which is the solution of

AðDÞgðtÞ ¼ BðDÞδðtÞ
gð0Þ ¼ 0; Dgð0Þ ¼ 0; . . . ; Dn1 gð0Þ ¼ 0

with δðtÞ being the Dirac delta function. Physically, gðtÞ is the response of the system under
an impulse of unit amplitude, which is often used to describe collision, impact, and shock. It
is easy to show that the impulse response function is related to the system transfer function
by

GðsÞ ¼ LfgðtÞg (7.1.12)

where L is the Laplace transform operator.

Thus, the free and force responses of a dynamic system can be obtained either by direct
inverse Laplace transform, as shown in Eqs. (7.1.5a) and (7.1.5b), or by the homogeneous
solution of Eq. (7.1.9) and the convolution integral of Eq. (7.1.11). These analytical methods
shall be used alternatively for convenience in analysis.

Transient Response and Steady-State Response

The response (output) yðtÞ of a dynamic system can be viewed as a sum of a transient
response ytr ðtÞ and a steady-state response yss . The transient response ytr ðtÞ describes the
change and transition of the system with respect to time and it is a function of time t. If
506 7 System Response Analysis

a system is stable and is under a bounded input, its transient response disappears after a long
enough time. (The stability of a dynamic system is examined in Section 7.2.) The steady-state
response yss is the remaining part of the system response after all transients have died out and
it is the time limit

yss ¼ lim yðtÞ (7.1.13)

t→∞

The yss is also known as a final value of the output. A transient response can be excited by
initial disturbances or by the application of an input. A steady-state response, however, can
only be excited by an input as it physically represents a balance between input energy and
output energy. In other words, a steady-state response is contained in the forced response
yF ðtÞ.
It should be pointed out that a steady-state response may not exist if the system is unstable
or if the input is unbounded. Figure 7.1.1 shows three cases of system responses: (a) response
y settles to a steady-state value yss after the transients die out; (b) response y has no steady
state because the limit lim yðtÞ does not exist; and (c) response y, being unbounded, has no
t→∞
steady state.

y
Transient response

yss

t
(a)

y y

t t
(b) (c)

Figure 7.1.1 Three cases of system response (see text for details)
 7.1 System Response Analysis in the Time Domain 507

Example 7.1.2
Consider a system with response y governed by

y_ þ σy ¼ r0 ; with yð0Þ ¼ y0

where σ > 0 and r0 is a constant. The solution of the differential equation by any method given in
Chapter 2 yields
 r0  σt r0
yðtÞ ¼ y0  e þ
σ σ

On the right-hand side of the previous equation, the first term is the transient response, which
vanishes as t goes to infinity; the second term is the steady-state response:
r0
yss ¼ lim yðtÞ ¼ :
t→∞ σ

Final-Value Theorem for Steady-State Response

The steady-state response of a dynamic system can be determined through direct use of
system response in the s-domain. According to the previous discussion, a steady-state
response can only be contained in the forced response yF ðtÞ, which in the s-domain is

YðsÞ ¼ GðsÞRðsÞ (7.1.14)

where Eq. (7.1.3) with IðsÞ ¼ 0 and Eq. (7.1.6) have been used. If all the poles of sYðsÞ have
negative real parts, by the final-value theorem described in Section 2.4.3, the steady-state
response of the system exists, and its final value is given by

yss ¼ lim yðtÞ ¼ lim sYðsÞ (7.1.15)

t→∞ s→0

¼ lim sGðsÞRðsÞ
s→0

When using Eq. (7.1.15), it is extremely important to check the condition of negative real parts
for all the poles of sY(s) of the final-value theorem.
Equation (7.1.15) delivers yss without the need to perform inverse Laplace transform and
to take the time limit, which is convenient in transfer function formulation.
508 7 System Response Analysis

Example 7.1.3
Consider the same system in Example 7.1.2. The s-domain response of the system is
r0
YðsÞ ¼ , which is obtained via the Laplace transform of the differential equation for the
sðs þ σÞ
r0
system with y0 ¼ 0. Thus, sYðsÞ ¼ has a pole p1 ¼ σ < 0, satisfying the condition of the
sþσ
final value theorem. It follows that the steady-state response of the system is
r0 r0
yss ¼ lim ¼
s→0 s þ σ σ

which is the same as the result obtained in Example 7.1.2.

System Order and Input Type

For convenience in analysis and for quantification of system response, system order and
input type are specified. The order of a dynamic system modeled by the transfer function in
Eq. (7.1.6) is the order n of the transfer function, which is also the order of its denominator A
(s) or the order of the relevant differential equation. Thus, G(s) in Eq. (7.1.6) is an n-th-order
transfer function.
Some of commonly used forcing functions are listed as follows:
Impulse function: rðtÞ ¼ I0 δðtÞ
Step function: rðtÞ ¼ r0
Ramp function: rðtÞ ¼ αt
Sinusoidal function: rðtÞ ¼ A sin ωt þ B cos ωt
where δðtÞ is the Dirac delta function, which is described in Section 2.4.1. Sinusoidal forcing
functions are mostly considered in frequency response analysis; see Section 7.3.

7.1.2 Time Responses of First-Order Systems

First-order systems are seen in various applications, including mass–damper systems, rotat-
ing shafts, resistor–capacitor (RC) and inductor–resistor (LR) circuits, liquid-level systems,
and thermal systems. A typical first-order system is governed by the differential equation

_ þ byðtÞ ¼ f ðtÞ
ayðtÞ (7.1.16)

where f ðtÞ and yðtÞ are the input and output, respectively. For example, the current of an LR
circuit is governed by

diðtÞ
L þ RiðtÞ ¼ vi ðtÞ
dt
 7.1 System Response Analysis in the Time Domain 509

where vi ðtÞ is an input voltage. If b = 0, Eq. (7.1.16) can be solved via direct integration:

ðt
1
yðtÞ ¼ yð0Þ þ f ðτÞdτ
a
0

Thus, in the subsequent discussion, we will consider b ≠ 0, which is seen in many applications.
For convenience in analysis, Eq. (7.1.16) is converted to a standard form

_ þ yðtÞ ¼ rðtÞ
T yðtÞ (7.1.17)
a 1
where T ¼ and rðtÞ ¼ f ðtÞ. The parameter T is known as the time constant of the system.
b b
The transfer function of the system is

YðsÞ 1
GðsÞ ¼ ¼ (7.1.18)
RðsÞ Ts þ 1

Free Response
The free response of a first-order system is governed by

_ þ yðtÞ ¼ 0;
T yðtÞ yð0Þ ¼ y0 (7.1.19)

By either the Laplace transform shown in Eq. (7.1.5a) or the formula in Eq. (7.1.9), the
solution of Eq. (7.1.19) is obtained as


t
T
yðtÞ ¼ y0 e (7.1.20)

Thus, the system free response decays exponentially. The value of the time constant
T determines the rate of decay of yðtÞ. The smaller the T, the faster the free response yðtÞ
decays, as shown in Figure 7.1.2.

Forced Response
The forced response of the first-order system is described by

_ þ yðtÞ ¼ rðtÞ;
T yðtÞ yð0Þ ¼ 0 (7.1.21)

By Eqs. (7.1.5b) and (7.1.18), the forced response is given by


1 1
yðtÞ ¼ L RðsÞ (7.1.22)
Ts þ 1
510 7 System Response Analysis

y (t )

T1 > T2
y0

T1

T2

Figure 7.1.2 Free-response curves of two first-order systems with different time constants

Now consider the forced response of a standard first-order system subject to the inputs
specified in Section 7.1.1.

Impulse Response
For a system under an impulse input rðtÞ ¼ I0 δðtÞ, its response is

 t
1 1 T
yðtÞ ¼ L1 I0 ¼ I0 gðtÞ; with gðtÞ ¼ e (7.1.23)
Ts þ 1 T

where gðtÞ is the impulse response of the system. The final value yss of the response is zero.
The impulse response curves of two systems with different time constants (T1 ¼ 2 and
T2 ¼ 1) are plotted in Figure 7.1.3 with MATLAB. As seen from the figure, a smaller time
constant leads to the response reaching zero more quickly.
It is seen from Figures 7.1.2 and 7.1.3 that the impulse response and the free response are
similar in pattern. This can be explained as follows. The impulse response is governed by

_ þ yðtÞ ¼ I0 δðtÞ; for t > 0

T yðtÞ
with yð0Þ ¼ 0

Integration of the previous equation in the domain 0  ≤ t ≤ 0þ

ð
0þ
  ð
0þ

_ þ yðtÞ dt ¼
T yðtÞ I0 δðtÞ dt ¼ I0
0 0

yields
Tyð0þÞ ¼ Tyð0Þ þ I0 ¼ I0
 7.1 System Response Analysis in the Time Domain 511

Impulse Response
1
T1
0.9 T2
0.8

0.7

0.6
y(t)

0.5

0.4

0.3

0.2

0.1

0
0 2 4 6 8 10
t (seconds)

Figure 7.1.3 Impulse response curves of two first-order systems of different time constants: T1 > T2 ,
generated by MATLAB
Ð
0þ
where, by the continuity of yðtÞ, yðtÞdt ¼ 0. Thus, the original impulse response problem is
0
converted to the following free-response problem

_ þ yðtÞ ¼ 0; for t > 0þ

T yðtÞ
with yð0þÞ ¼ I0 =T

This explanation can be extended to general linear time-invariant systems.

Step Response
For a system under a step input rðtÞ ¼ r0 , its response is
0 1
 t
1 r0 
yðtÞ ¼ L1 ¼ r0 @ 1  e T A (7.1.24)
Ts þ 1 s

The final value of the system response (the steady-state response) is yss ¼ r0 : Figure 7.1.4
shows the step-response curves of two systems with different time constants. As can be seen
from the figure, a smaller time constant results in the response reaching the final value more
quickly.

Ramp Response
For a system under a ramp forcing function, rðtÞ ¼ αt, its response is
512 7 System Response Analysis

y (t )

yss
r0

T2

T1

T1 > T2

Figure 7.1.4 Step-response curves of two first-order systems with different time constants

8 0 19
 < t =
1 α 
yðtÞ ¼ L1 ¼ α t  T @1  e T A (7.1.25)
Ts þ 1 s2 : ;

As time goes by, the forced response becomes unbounded, yðtÞ → αðt  TÞ. Equation (7.1.25)
also implies that the response lags the input by a time T.
Now define an error function as the difference between the input and response
0 1
t

eðtÞ ≡ rðtÞ  yðtÞ ¼ αT @1  e T A (7.1.26)

According to Eq. (7.1.26), the steady-state error exists:

ess ¼ lim eðtÞ ¼ αT (7.1.27)

t→∞

See Figure 7.1.5 for the input and response curves, time lag, and steady-state error.

Response to General Inputs

The impulse response function of a system by Eq. (7.1.12) is

 t
1 1 1 
gðtÞ ¼ L ¼ e T (7.1.28)
Ts þ 1 T

which is also shown in Eq. (7.1.23). The forced response of the system subject to an
arbitrary input can be obtained by the convolution integral of Eq. (7.1.11), and it is
given by
 7.1 System Response Analysis in the Time Domain 513

y (t )

αT

T
r (t )

y (t )

Figure 7.1.5 Ramp response and steady-state error of a first-order system

ðt ðt tτ
1 
yðtÞ ¼ gðt  τÞrðτÞdτ ¼ e T rðτÞdτ (7.1.29)
T
0 0

Example 7.1.4
1
Under an input rðtÞ ¼ r0 ð1  eσt Þ, with σ > 0 and σ ≠ , the system response by Eq. (7.1.29) is
T
determined as

ðt tτ
1 
yðtÞ ¼ e T r0 ð1  eστ Þdτ
T
0
2 3τ¼t
t τ 1
r0  T ð1  σTÞτ
¼ e T 4TeT  eT 5
T 1  σT
0 1τ¼0
t
σT  1
¼ r0 þ r0 @ e T eσt A
1  σT 1  σT

On the right-hand side of the previous equation, the first term is the steady-state response, yss ¼ r0 ;
the second term is the transient response, which vanishes as time goes by.

Estimation of the Time Constant

As shown in Figures 7.1.2 to 7.1.5, the time constant T of a first-order system characterizes its
free and forced responses. Hence, modeling and analysis of a first-order system requires
the knowledge of its time constant. The time constant of a first-order system can be estimated
514 7 System Response Analysis

from its response curves. For instance, consider the step response shown in Figure 7.1.6. At
 
t = T, the system response by Eq. (7.1.24) is yðTÞ ¼ r0 1  e1 ¼ 0:632yss . In other words,
T is the time at which the step response is 63.2% of the steady-state value. Also, the slope of
1 yss
_
the step response at the initial time is yð0Þ ¼ r0 , indicating that T ¼ . Additionally, the
T _
yð0Þ
steady-state error shown in Figure 7.1.5 can be used to estimate the time constant. Indeed,
Eq. (7.1.27) leads to T ¼ ess =α.

y (t )

yss
r0

0.632yss
r0
slope =
T

t
T

Figure 7.1.6 Step response at t = T and slope at t = 0

Example 7.1.5
The output of a first-order system subject to a step input has a final value yss ¼ 2:5. At t = 2, the
system response is 1.0. Determine the time constant of the system.

Solution
By Eq. (7.1.24),
0
1
2

yð2Þ ¼ 2:5  @1  e T A ¼ 1

where yss ¼ r0 has been used. Thus,

2
 1
e T ¼1 ¼ 0:6
2:5

It follows that the time constant of the system is

2
T¼ ¼ 3:915
lnð0:6Þ
 7.1 System Response Analysis in the Time Domain 515

7.1.3 Time Responses of Second-Order Systems

Second-order systems are seen in many applications, including spring–mass–damper sys-
tems, rotating shafts with torsional springs and dampers, RCL circuits, direct current (DC)
motors, liquid-level systems, thermal systems, and first-order systems with feedback control.
A typical second-order system is governed by the differential equation

a€ _ þ cyðtÞ ¼ f ðtÞ
y þ byðtÞ (7.1.30)

with a > 0; b ≥ 0; c > 0, where f ðtÞ and yðtÞ are the system input and output, respectively.
For convenience in analysis, Eq. (7.1.30) is converted to a standard form

y€ þ 2ξωn yðtÞ
_ þ ω2n yðtÞ ¼ ω2n rðtÞ (7.1.31)

where ωn and ξ are the natural frequency and damping ratio of the system, respectively, and
they are called model parameters. The model parameters are related to the physical param-
eters a, b, c by
rffiffiffi
c 1 b
ωn ¼ ; ξ ¼ pffiffiffiffiffi (7.1.32)
a 2 ac

1
and the scaled input rðtÞ ¼ f ðtÞ. In addition, the following two parameters are often used in
c
analysis:
ωn
Natural frequency fn ¼ ; in Hz
2π
1 2π
Natural period T¼ ¼ ; in seconds
fn ω n

The transfer function of the system, from Eq. (7.1.31), is

YðsÞ ω2n
GðsÞ ¼ ¼ 2 (7.1.33)
RðsÞ s þ 2ξωn s þ ω2n

The poles of the transfer function are the roots of the characteristic equation

s2 þ 2ξωn s þ ω2n ¼ 0 (7.1.34)

and they are

qffiffiffiffiffiffiffiffiffiffiffiffiffi
s1
¼ ξωn  ξ 2  1 ωn (7.1.35)
s2
516 7 System Response Analysis

Example 7.1.6
The equation of motion of a spring–mass–damper system is

_ þ kxðtÞ ¼ f ðtÞ
x þ cxðtÞ
m€

which can be converted to Eq. (7.1.31), with

rffiffiffiffi
k 1 c 1
ωn ¼ ; ξ ¼ pffiffiffiffiffiffi ; rðtÞ ¼ f ðtÞ:
m 2 mk k

The transfer function of the system in terms of the physical parameters is

X ðsÞ 1
¼
FðsÞ ms2 þ cs þ k

which can be converted to the standard form of Eq. (7.1.33) if the scaled input rðtÞ is used. The
transfer function poles are the roots of the characteristic equation ms2 þ cs þ k ¼ 0, which are
obtained as

s1 1  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ c  c2  4mk
s2 2m

A more general form of second-order systems is

a€ _ þ cyðtÞ ¼ αf_ ðtÞ þ βf ðtÞ

y þ byðtÞ (7.1.36)

The transfer function for such a system is

YðsÞ ðαs þ βÞ ω2n

¼ 2 (7.1.37)
RðsÞ s þ 2ξωn s þ ω2n

if the scaled input rðtÞ ¼ f ðtÞ=c is used. It is observed that the transfer function has a
zero s ¼ β=α, which induces numerator dynamics. Also, a comparison of the transfer
functions in Eqs. (7.1.33) and (7.1.37) shows that the two systems have the same characteris-
tic equation (7.1.34). This indicates that the input f(t) does not alter the transfer function
poles.
Without loss of generality, the standard form of Eq. (7.1.33) will now be used to examine
the system response. The effect of the zero on the system response, as shown in the general
form of Eq. (7.1.36), shall be discussed in Section 7.1.3.
 7.1 System Response Analysis in the Time Domain 517

Four Damping States

According to Eq. (7.1.35), there are four cases or states of damping in the system.

(a) No damping (ξ ¼ 0)
The poles of the system are of the imaginary form

s1 pffiffiffiffiffiffiffi
¼  j ωn ; j ¼ 1 (7.1.38)
s2

In this case, the system is undamped, and it is called an undamped system.

(b) Underdamping (0 < ξ < 1)

The poles of the system are a complex conjugate pair with negative real part:

s1 pffiffiffiffiffiffiffi
¼ σ  jωd ; j ¼ 1 (7.1.39)
s2
pffiffiffiffiffiffiffiffiffiffiffiffiffi
with σ ¼ ξωn and ωd ¼ 1  ξ 2 ωn . Here σ and ωd are the decay factor and damped
frequency, respectively. In this case, the system is underdamped, and it is called an under-
damped system.

(c) Critical damping (ξ ¼ 1)

The poles of the system are identical and they are real and negative as follows:

s1
¼ ωn (7.1.40)
s2

In this case, the system is critically damped, and it is called a critically damped system.
Critical damping is the border between underdamping and overdamping cases.

(d) Overdamping (ξ > 1)

The poles of the system are two distinct real and negative numbers, as given by (7.1.35), and
they can be written as

s1
¼ σ  β (7.1.41)
s2
pffiffiffiffiffiffiffiffiffiffiffiffiffi
with σ ¼ ξωn and β ¼ ξ 2  1 ωn . In this case, the system is overdamped, and it is called an
overdamped system.
The above four damping cases are applicable to a second-order system of the general form
of Eq. (7.1.37) because it has the same characteristic equation (Eq. (7.1.34)).
518 7 System Response Analysis

The four damping cases of second-order systems show that the value of the damping ratio
ξ determines the pole locations in the complex plane; see Figure 7.1.7, where crosses ðÞ
indicate the pole locations. As shown in the figure, the poles of the system in cases (b) to (d)
all have negative real parts. As shall be seen subsequently, different pole locations give
different patterns of system response.

(a) (b)

(c) (d)

Figure 7.1.7 Damping ratio and pole locations: cross ¼ pole location

Example 7.1.7
For a spring–mass–damper system, m€ _ þ kxðtÞ ¼ f ðtÞ, derive the conditions on the system
x þ cxðtÞ
parameters (m, c, k) such that the system is underdamped.

Solution
1 c
From Example 7.1.6, the damping ratio is given by ξ ¼ pffiffiffiffiffiffi. For underdamping, 0 < ξ < 1,
2 mk
1 c
which requires that 0 < pffiffiffiffiffiffi < 1. Thus, the conditions of underdamping are
2 mk
c > 0 and c2 < 4mk

Another way to solve this problem is to consider the locations of the transfer function poles,
without having to derive the damping ratio in terms of the system parameters. From
Example 7.1.6, the characteristic equation of the system is ms2 + cs + k = 0, from which the
poles are

s1 1  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ c  c2  4mk
s2 2m

For the system to be underdamped, its poles s1 and s2 must be a complex-conjugate pair with
a negative real part (Figure 7.1.7b). This leads to the conditions of c > 0 and c2 < 4mk.
 7.1 System Response Analysis in the Time Domain 519

Example 7.1.8
Given a system

X ðsÞ 8
¼
FðsÞ s2 þ 3s þ 8

compute its natural frequency and damping ratio, and comment on the damping characteristics of
the system.
Solution
Comparing the coefficients of the characteristic equation for the system

s2 þ 3s þ 8 ¼ 0

with those of Eq. (7.1.34), yields

2ξωn ¼ 3; ω2n ¼ 8

Solution of the previous algebraic equations gives

pffiffiffi
pffiffiffi 3 2
ωn ¼ 2 2 ¼ 2:8284; ξ ¼ ¼ 0:5303
8

Because 0 < ξ < 1, the system is underdamped.

Free Response
The free response of a second-order system is governed by

_ þ ω2n yðtÞ ¼ 0
y€ þ 2ξωn yðtÞ (7.1.42)

_
with the initial conditions yð0Þ ¼ y0 ; yð0Þ ¼ v0 . Substituting y ¼ Aeλt into the homogeneous
equation yields the characteristic equation

λ2 þ 2ξωn λ þ ω2n ¼ 0 (7.1.43)

Through comparison of Eq. (7.1.34) and Eq. (7.1.43), it is concluded that the poles s1 ; s2 of
the transfer functions are the same as the characteristic roots λ1 ; λ2 for the homogeneous
solution. Thus, the free-response solution can be expressed as

yðtÞ ¼ A1 es1 t þ A2 es2 t (7.1.44)

with A1 and A2 to be determined by the initial conditions.

520 7 System Response Analysis

The free response of a system is determined in the following four damping cases.

(a) No damping ðξ ¼ 0Þ
The homogeneous solution can be written as

yðtÞ ¼ A1 e jωn t þ A2 ejωn t

¼ B1 cos ωn t þ B2 sin ωn t

where Euler’s formula e jθ ¼ cos θ þ j sin θ has been used. The constants B1 and B2 can be
directly determined by the initial conditions, without the need to know A1 and A2 . It follows
that the free response is given by
v0
yðtÞ ¼ y0 cos ωn t þ sin ωn t (7.1.45)
ωn

(b) Underdamping ð0 < ξ < 1Þ

The homogeneous solution is given by

yðtÞ ¼ A1 eðσþjωd Þt þ A2 eðσjωd Þt

¼ eσt ðB1 cos ωd t þ B2 sin ωd tÞ
pffiffiffiffiffiffiffiffiffiffiffiffiffi
with σ ¼ ξωn and ωd ¼ 1  ξ 2 ωn . From the initial conditions, the free response is
obtained as

σt v0 þ σy0
yðtÞ ¼ e y0 cos ωd t þ sin ωd t (7.1.46)
ωd

(c) Critical damping ðξ ¼ 1Þ

Because the system has repeated poles as shown in Eq. (7.1.40), Eq. (7.1.44) is not usable.
From differential-equation theory, the solution of Eq. (7.1.42) in this case is of the form

yðtÞ ¼ eωn t ðB1 t þ B0 Þ

Thus, with the initial conditions, the free response is determined as follows:
 
yðtÞ ¼ eωn t y0 þ ðv0 þ ωn y0 Þt (7.1.47)

(d) Overdamping (ξ > 1)

The homogeneous solution is given by

yðtÞ ¼ A1 eðσþβÞt þ A2 eðσβÞt

¼ eσt ðB1 cosh βt þ B2 sinh βtÞ
pffiffiffiffiffiffiffiffiffiffiffiffiffi
with σ ¼ ξωn and β ¼ ξ 2  1 ωn . From the initial conditions, the free response is
obtained as
 7.1 System Response Analysis in the Time Domain 521

y (t ) y (t )

t t

ξ=0 0 < ξ <1

y (t ) y (t )

t t

ξ=1 ξ >1

Figure 7.1.8 The influence of the damping ratio ξ on the free response of a second-order system

Free Response to y0 = 1 and v0 = 0

1
ξ=0
0.8 ξ = 0.3
ξ=1
0.6 ξ = 1.5

0.4

0.2
y(t)

–0.2

–0.4

–0.6

–0.8

–1
0 0.5 1 1.5 2 2.5 3
t

Figure 7.1.9 Free responses with ωn ¼ 4 and ξ ¼ 0; 0:3; 1:0; and 1:5, subject to y0 ¼ 1; v0 ¼ 0, plotted
with MATLAB


σt v0 þ σy0
yðtÞ ¼ e y0 cosh βt þ sinh βt (7.1.48)
β

The free-response curves of a second-order system in the four damping cases are shown in
Figure 7.1.8. As can be seen from the figure, the damping ratio of a system has significant influence
on the pattern of the free response: from oscillation to decayed oscillation to exponential decay, as
ξ increases.
522 7 System Response Analysis

To see the effects of damping in further detail, consider four systems with the same natural
frequency ωn ¼ 4 and different damping ratios ξ ¼ 0; 0:3; 1:0; and 1:5: Assume the same
_
initial conditions for all the systems: yð0Þ ¼ 1 and yð0Þ ¼ 0. The free-response curves are
computed from Eqs. (7.1.45) to (7.1.48), and they are plotted in Figure 7.1.9. As seen from
the figure, in the first half cycle, the undamped system ðξ ¼ 0Þ takes a shorter time
ðt ¼ π=ð2ωn Þ ¼ 0:393sÞ to reach the equilibrium position (y = 0) than the underdamped
system ðξ ¼ 0:3Þ. The critically damped system ðξ ¼ 1:0Þ and the overdamped system
ðξ ¼ 1:5Þ, on the other hand, never reach the equilibrium position because their responses
are nonoscillatory.

Forced Response
The forced response of a second-order system is governed by the differential equation

y€ þ 2ξωn yðtÞ
_ þ ω2n yðtÞ ¼ rðtÞ (7.1.49)

_
with the zero initial conditions yð0Þ ¼ 0; yð0Þ ¼ 0. The solution of Eq. (7.1.49) by the
transfer function formulation (7.1.5b), is

1 ω2n
yðtÞ ¼ L RðsÞ (7.1.50)
s2 þ 2ξωn s þ ω2n

where the transfer function given in Eq. (7.1.33) has been used. Refer to Section 2.4 for
inverse Laplace transforms.
We now examine the impulse and step responses of second-order systems.

Response to an Impulse Input

For a system under an impulse input rðtÞ ¼ I0 δðtÞ; its forced response is

yðtÞ ¼ I0 gðtÞ (7.1.51)

where gðtÞ is the impulse response function given by

gðtÞ ¼ L1 fGðsÞg


1 ω2n
¼L (7.1.52)
s2 þ 2ξωn s þ ω2n

The impulse response function is obtained in the following four damping cases.

(a) Undamped system ðξ ¼ 0Þ

gðtÞ ¼ ωn sin ωn t (7.1.53a)

 7.1 System Response Analysis in the Time Domain 523

(b) Underdamped system ð0 < ξ < 1Þ

ω2n σt ωn
gðtÞ ¼ e sin ωd t ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi eσt sin ωd t (7.1.53b)
ωd 1  ξ2

(c) Critically damped system ðξ ¼ 1Þ

gðtÞ ¼ ω2n t eωn t (7.1.53c)

(d) Overdamped system ðξ > 1Þ

ω2n σt ωn
gðtÞ ¼ e sinh βt ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi eσt sinh βt (7.1.53d)
β ξ2  1
pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi
In the above formulas, σ ¼ ξωn , ωd ¼ 1  ξ 2 ωn and β ¼ ξ 2  1 ωn .
In Case (b) to iv ðξ > 0Þ, the steady-state response of the system is zero. The steady-
state response in Case (a) ðξ ¼ 0Þ does not exist.
The impulse response function g(t) in Eqs. (7.1.53a–d) can be obtained by the
inverse Laplace transform of the system transfer function G(s). In Cases (a), (b), and
(d), G(s) has two distinct poles p1 and p2 , and it can be expressed by partial fraction
expansion:

K1 K2
GðsÞ ¼ þ
s  p1 s  p2

In Case (c), G(s) has two identical poles, p1 ¼ p2 ¼ p, and it can be written as

b1 b2
GðsÞ ¼ þ
s  p ðs  pÞ2

Here, the coefficients K1 , K2 , b1 , and b2 are the residues of G(s). Thus, determination of the
transfer function residues gives the impulse response function as follows:

K1 ep1 t þ K2 ep2 t for Cases ðaÞ; ðbÞ; and ðdÞ

gðtÞ ¼
ept ðb1 þ b2 tÞ for Case ðcÞ

The curves of the impulse response function gðtÞ are plotted in Figure 7.1.10 for the four
damping cases: ωn ¼ 4 and ξ ¼ 0; 0:3; 1:0; 1:5:
524 7 System Response Analysis

2
ξ=0
1.5 ξ = 0.3
ξ=1
ξ = 1.5
1

0.5
g(t)

–0.5

–1

–1.5

–2
0 0.5 1 1.5 2 2.5 3
t

Figure 7.1.10 Impulse response function g(t) with ωn ¼ 4 and ξ ¼ 0; 0:3; 1:0 and 1:5, plotted with
MATLAB

Example 7.1.9
X ðsÞ 8
Given a system ¼ , determine its forced response subject to an impulse input
FðsÞ s2 þ 3s þ 8
rðtÞ ¼ 5δðtÞ.

Solution
From Example 7.1.8, the damping ratio of the system is ξ ¼ 0:5303, indicating that the system is
underdamped. Also, by the characteristic equation

s2 þ 3s þ 8 ¼ 0

the poles of the transfer function are

p1 1 pffiffiffiffiffi
¼ ð3  j 23Þ ¼ 1:5  j2:3979
p2 2
which gives σ ¼ 1:5; ωd ¼ 2:3979. It follows from Eqs. (7.1.51) and (7.1.53b) that the impulse
response of the system is given by

1
yðtÞ ¼ 5  e1:5t sin 2:3979t
2:3979
¼ 2:0851 e1:5t sin 2:3979t
 7.1 System Response Analysis in the Time Domain 525

Response to a Step Input

For a system subject to a step input rðtÞ ¼ r0 ; its forced response is given by
n 
r0 o ω2n r0
yðtÞ ¼ L1 GðsÞ ¼ L1 2 (7.1.54)
s s þ 2ξωn s þ ω2n s

The inverse Laplace transform in Eq. (7.1.26) can be carried out through the use of partial
fraction expansion of the system transfer function (Section 2.4):

ω2n r0 K0 K1 K2
¼ þ þ
s2 þ 2ξωn s þ ωn s
2 s s  p1 s  p2

where p1 and p2 are the poles of G(s), as given in Eq. (7.1.35). If the poles are repeated (as in
the case of critical damping), p1 ¼ p2 ¼ p and,

ω2n r0 K0 b1 b2
¼ þ þ
s þ 2ξωn s þ ωn s
2 2 s s  p ðs  pÞ2

With the above expressions of partial fraction expansion, the step response of a second-
order system is obtained in the following four damping cases.

(a) Undamped system ðξ ¼ 0Þ

yðtÞ ¼ r0 ð1  cos ωn tÞ (7.1.55a)

(b) Underdamped system ð0 < ξ < 1Þ

 
σ
xðtÞ ¼ r0 1  eσt cos ωd t þ sin ωd t (7.1.55b)
ωd

(c) Critically damped system ðξ ¼ 1Þ

yðtÞ ¼ r0 f1  eωn t ð1 þ ωn tÞg (7.1.55c)

(d) Overdamped system ðξ > 1Þ

 
σ
yðtÞ ¼ r0 1  eσt cosh βt þ sinh βt (7.1.55d)
β
pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi
In the above formulas, σ ¼ ξωn , ωd ¼ 1  ξ 2 ωn and β ¼ ξ 2  1 ωn . In Cases (b) to (d)
ðξ > 0Þ, the system response has a final value yss ¼ r0 : The steady-state response in Case (a)
ðξ ¼ 0Þ does not exist.
526 7 System Response Analysis

Step Response
2

1.8 ξ = 0.1

1.6

1.4 0.3

1.2 0.5
0.7
y(t)

1
1.0
0.8 1.5

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3
t

Figure 7.1.11 Step responses with ωn ¼ 4 and ξ ¼ 0:1; 0:3; 0:5; 0:7; 1:0; 1:5, subject to unit step input
r0 ¼ 1

Figure 7.1.11 shows the curves of the step responses of second-order systems with ωn ¼ 4
and ξ ¼ 0:1; 0:3; 0:5; 0:7; 1:0; 1:5, subject to the unit step input r0 ¼ 1. The effect of
damping on the system response is observed. With underdamping ðξ ¼ 0:1; 0:3; 0:5; 0:7Þ,
the system response is oscillatory; with critical damping ðξ ¼ 1:0Þ and over damping
ðξ ¼ 1:5Þ, the system is nonoscillatory. All the response curves approach the same steady-
state value, yss ¼ 1.

General Second-Order Systems

For a second-order system in the general form of Eq. (7.1.37), its free response is the same as
a system in the standard form because the two systems have the same governing equation
(7.1.42). The forced response of a general second-order system in the s-domain can be written as

ðαs þ βÞ ω2n
YðsÞ ¼ RðsÞ
s2 þ 2ξωn s þ ω2n
(7.1.56)
ω2n
¼ 2 fαR1 ðsÞ þ βR2 ðsÞg
s þ 2ξωn s þ ω2n

with R1 ðsÞ ¼ sRðsÞ and R2 ðsÞ ¼ RðsÞ. This indicates that the forced response of a
general second-order system can be viewed as the response of the standard system subject
to an input that is the linear combination of R1 ðsÞ and R2 ðsÞ. For instance, if a general system
 7.1 System Response Analysis in the Time Domain 527

is subject to a step input rðtÞ ¼ r0 , r1 ðtÞ ¼ r0 δðtÞ and r2 ðtÞ ¼ r0 . Thus, the step response yðtÞ of
a general system is a linear combination of the impulse response zimp ðtÞ and the step response
zstp ðtÞ of the standard system:

yðtÞ ¼ α zimp ðtÞ þ β zstp ðtÞ (7.1.57)

where

1 ω2n
zimp ðtÞ ¼ L r0
s2 þ 2ξωn s þ ω2n

ω2n r0
zstp ðtÞ ¼ L1
s þ 2ξωn s þ ωn s
2 2

Also, by the final-value theorem in Section 7.1.1, the step response has a final value yss ¼ βr0 .

Example 7.1.10

X ðsÞ 4s þ 1
For a system ¼ subject to a step input rðtÞ ¼ 3, obtain a mathematical expression
FðsÞ s2 þ 3s þ 8
of the forced response and plot the response.

Solution
Using Eq. (7.1.56)

4s þ 1 3 8 3 8 3
YðsÞ ¼ ¼ þ (a)
s2 þ 3s þ 8 s s2 þ 3s þ 8 2 s2 þ 3s þ 8 8s

From Examples 7.1.8 and 7.1.9, the system is underdamped, with

pffiffiffi
pffiffiffi 3 2
ωn ¼ 2 2 ¼ 2:8284; ξ ¼ ¼ 0:5303
8
qffiffiffiffiffiffiffiffiffiffiffiffiffi
σ ¼ ξωn ¼ 1:5; ωd ¼ 1  ξ 2 ωn ¼ 2:3979

Thus, application of Eqs. (7.1.53b) and (7.1.55b) to Eq. (a) yields the step response of the system as
follows:
 
3 ω2n σt 3 σt σ
yðtÞ ¼ e sin ωd t þ 1e cos ωd t þ sin ωd t
2 ωd 8 ωd
3
¼ 5:0043e1:5t sin 2:3979t þ 1  e1:5t ðcos 2:3979t þ 0:6255 sin 2:3979tÞg (b)
8

By Eq. (b), the step response is plotted in Figure 7.1.12. As can be seen from the figure, the step
response approaches a steady-state value yss ¼ 3=8 ¼ 0:375.
528 7 System Response Analysis

Figure 7.1.12 The step response of the system in Example 7.1.10

Note that the step response in Figure 7.1.12 can be easily generated by the MATLAB function
step, with the transfer function formulation given in Eq. (a). This is done by the following
MATLAB commands

sys = tf(3*[0 4 1], [1 3 8])

[y, t] = step(sys, 5);
plot(t, y, ‘k’, ‘LineWidth’, 2);
grid; xlabel(‘t’); ylabel(‘y(t)’)

7.1.4 Specification of Step Responses of Underdamped Second-Order Systems

Step responses of underdamped second-order systems are commonly considered in dynamic
analysis and feedback control in engineering applications. For an underdamped system
ð0 < ξ < 1Þ, its step response by Eq. (7.1.55b) is
 
σt σ
yðtÞ ¼ r0 1  e cos ωd t þ sin ωd t
ωd
( qffiffiffiffiffiffiffiffiffiffiffiffiffi )
1
¼ r0 1  pffiffiffiffiffiffiffiffiffiffiffiffiffi eξωn t cos 1  ξ 2 ωn t  ϕ (7.1.58)
1  ξ2
 7.1 System Response Analysis in the Time Domain 529

Figure 7.1.13 Pole plot of an underdamped system in the complex plane

pffiffiffiffiffiffiffiffiffiffiffiffiffi
where ϕ ¼ tan1 ðξ= 1  ξ 2 Þ: Step-response curves for some values of ξ are plotted in
Figure 7.1.11.

Graphic Interpretation of Model Parameters

To better understand the influence of the model parameters ωn and ξ on the step response,
the pole s ¼ σ þ jωd of an underdamped system in the complex plane is shown in
pffiffiffiffiffiffiffiffiffiffiffiffiffi
Figure 7.1.13, where from Eq. (7.1.39) σ ¼ ξωn and ωd ¼ 1  ξ 2 ωn . From the figure, it
is easy to show that the magnitude and phase angle of the pole are related to the model
parameters by
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
jsj ¼ σ2 þ ω2d ¼ ωn
ω  (7.1.59)
∠s ¼ tan1
d
¼πθ
σ

where θ ¼ tan1 ðωd =σÞ, in rad. It follows that

σ
cos θ ¼ ¼ξ (7.1.60)
jsj

Two observations can be made on the above results: (i) the further away the pole s is from
the origin, the larger the natural frequency ωn ; and (ii) the smaller the angle θ, the larger the
damping ratio ξ. Also, it is seen from Eq. (7.1.60) that overdamped and undamped cases
occur when θ is zero and π/2, respectively.

Performance Specification Parameters

To quantify the performance of an undamped second-order system, its step response
described by Eq. (7.1.58) is specified in Figure 7.1.14, where tr , tp , Mp , and ts are the rise
530 7 System Response Analysis

y (t )

ymax

yss
MP × yss

r0

− δ × yss
+

t
tr tp ts

Figure 7.1.14 Specification of a step response

time, peak time, maximum overshoot, and settling time, respectively. These performance
specification parameters are defined as follows:

(a) Rise time tr , the time required for the output yðtÞ to rise from 0% to 100% of its final
value, yss ¼ r0 . There are other definitions of the rise time, such as the time required for
the output to go from 10% to 90% of yss . In this text, the 100% rise-time definition is used.
(b) Peak time tp , the time required for the output yðtÞ to reach the first and maximum peak;
that is, yðtp Þ ¼ ymax .
(c) Maximum overshoot Mp , a non-dimensional number specifying the amount of the output
yðtÞ over its final value yss at the peak time tp ; namely,

yðtp Þ  yss
Mp ¼ (7.1.61)
yss

In other words, yðtp Þ=yss ¼ 1 þ Mp . In applications, the maximum overshoot is often

specified in percentage as follows:

%OS ¼ Mp  100% (7.1.62)

Here %OS is called the percent maximum overshoot.

(d) Settling time ts , the time required for the output oscillations to stay with the specified
percentage ðδ  100%Þ of the final value yss . The commonly used values of δ  100%
are 1%, 2%, and 5%. Here δ specifies an error band or allowable tolerance.
 7.1 System Response Analysis in the Time Domain 531

The rise time, peak time, and settling time of a system tell how fast the system is in response
to a step input. The maximum overshoot is a measure of the oscillatory character of the
system response, and it is also an indicator of the relative stability of the system. These
performance specification parameters are useful in the design of second-order systems.
We now evaluate the performance specification parameters in terms of ωn and ξ.

Rise Time
According to the rise-time definition and from Eq. (7.1.58), we have
 
σ
yðtr Þ ¼ r0 1  eσt cos ωd tr þ sin ωd tr ¼ r0
ωd

which implies that

σ
cos ωd tr þ sin ωd tr ¼ 0
ωd

The solution of the previous trigonometry equation gives

1  
tr ¼ π  tan1 ðωd =σÞ
ωd  qffiffiffiffiffiffiffiffiffiffiffiffiffi
1 (7.1.63)
¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi π  tan1 ð 1  ξ 2 =ξÞ
1  ξ 2 ωn

Equation (7.1.63) indicates that for a given value of damping ratio ξ, the rise time is inversely
proportional to the natural frequency ωn . Note that π  tan1 ðωd =σÞ is the phase angle of the
pole s ¼ σ þ jωd and, from Eq. (7.1.59), the rise time can also be written as

πθ
tr ¼ (7.1.64)
ωd

where the angle θ is given by Eq. (7.1.60) and it is in radians.

Maximum Overshoot
The maximum overshoot occurs at the peak time, which satisfies dyðtp Þ=dt ¼ 0. From Eq.
(7.1.58),
 
dyðtÞ σt σ σt
¼ r0 σe cos ωd t þ sin ωd t  e ðωd sin ωd t þ σ cos ωd tÞ
dt ωd
1  2 
¼ r0 σ þ ω2d eσt sin ωd t ¼ 0
ωd

This implies that sin ωd tp ¼ 0, which has solutions ωd tp ¼ nπ; n ¼ 1; 2; . . . Because of the
exponentially decaying factor eσt in Eq. (7.1.58), yðtp Þ ¼ ymax occurs in the first cycle of
oscillation, n = 1 (see Figure 7.1.14). Thus,
532 7 System Response Analysis

π π
tp ¼ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi (7.1.65)
ωd 1  ξ 2 ωn

It follows from Eq. (7.1.58) that


yðtp Þ σ
¼ 1 þ Mp ¼ 1  eσtp cos ωd tp þ sin ωd tp ¼ 1 þ eσtp
r0 ωd

in which Eq. (7.1.65) has been used. Consequently,

pffiffiffiffiffiffiffi2ffi
Mp ¼ eσtp ¼ eπξ= 1ξ
(7.1.66)

Equation (7.1.66) shows that the maximum overshoot is only dependent on the damping
ratio, and that an increase in ξ decreases Mp .
If the maximum overshoot of a step response is known, either experimentally measured or
prescribed in design, the damping ratio ξ, according to Eq. (7.1.66), is related to Mp by
 2
ln Mp
ξ ¼
2
 2 (7.1.67)
π2 þ ln Mp

Settling Time
The oscillations of the step response about its final value ðyss ¼ r0 Þ are described by
qffiffiffiffiffiffiffiffiffiffiffiffiffi
yðtÞ  yss 1 ξωn t
¼  pffiffiffiffiffiffiffiffiffiffiffiffiffi e cos 1  ξ 2 ωn t  ϕ
yss 1  ξ2
1
where Eq. (7.1.58) has been used. Because jcos θj≤1 for any θ, the curves  pffiffiffiffiffiffiffiffiffiffiffiffiffi eξωn t
1  ξ2
define an envelope for the step response; see Figure 7.1.15, which also shows that the

oscillations after the settling time ts are confined in the error band




yðtÞ  yss
1


ξωn t

yss
≤ pffiffiffiffiffiffiffiffiffiffiffiffi2ffi e ≤ δ; for t ≥ ts
1ξ

Accordingly, the settling time with the δ  100% criterion can be estimated by

1
pffiffiffiffiffiffiffiffiffiffiffiffiffi eξωn ts ¼ δ (7.1.68)
1  ξ2

which gives
 qffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffi
1 1
ts ¼  ln δ 1  ξ 2 ¼ ln δ þ ln 1  ξ2 (7.1.69)
ξωn ξωn
g
 7.1 System Response Analysis in the Time Domain 533

y (t )
y(t )−yss
yss
e−ξ ωnt
1− ξ 2

t = ts
δ

0 t

−δ

e−ξ ωnt
−
1− ξ 2

Figure 7.1.15 Envelope of the step response and the settling time

pffiffiffiffiffiffiffiffiffiffiffiffiffi
In practice, jln δj is much larger than jln 1  ξ 2 j. Thus, the following simplified formula
for the settling time is often used

ln δ
ts ¼  (7.1.70)
ξωn
qffiffiffiffiffiffiffiffiffiffiffiffiffi
which is obtained from Eq. (7.1.69) by neglecting ln 1  ξ2 . Now, we calculate three
numbers: lnð0:05Þ ¼ 2:9957; lnð0:02Þ ¼ 3:9120; and lnð0:01Þ ¼ 4:6051: Rounding
these numbers and substituting the results into Eq. (7.1.70), yields three commonly used
formulas for estimation of the settling time:

3
5% criterion ðδ ¼ 0:05Þ : ts ¼
ξωn
4
2% criterion ðδ ¼ 0:02Þ : ts ¼ (7.1.71)
ξωn
4:6
1% criterion ðδ ¼ 0:01Þ : ts ¼
ξωn

Other criteria with different values of δ, of course, can be derived from Eq. (7.1.70). For
5:3
instance, for δ ¼ 0:5%, lnð0:005Þ ¼ 5:2983 and ts ¼ .
ξωn
534 7 System Response Analysis

Equations (7.1.70) and (7.1.71) are good for a relatively small damping ratio. If the
damping ratio is relatively large, say 0:7 ≤ ξ ≤ 0:9, Eq. (7.1.69) may be used to obtain
a modified formula:

1 n pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi o 1
ts ¼  ln δ þ ln 1  0:82 ¼  fln δ  0:5108g
ξωn ξωn
g

Here, 0.8 is the midpoint of the range 0:7 ≤ ξ ≤ 0:9. In this case, the settling time with a 2%
4:51
criterion becomes ts ¼ .
ξωn

Example 7.1.11
X ðsÞ 8
For an underdamped system ¼ , compute the rise time, maximum overshoot, and
FðsÞ s2 þ 3s þ 8
settling time with a 2% criterion of its step response.

Solution
The natural frequency and damping ratio of the system are
pffiffiffi
pffiffiffi 3 2
ωn ¼ 2 2 ¼ 2:8284; ξ ¼ ¼ 0:5303
8

which were obtained in Example 7.1.8. Thus,

qffiffiffiffiffiffiffiffiffiffiffiffiffi
σ ¼ ξωn ¼ 1:5; ωd ¼ 1  ξ 2 ωn ¼ 2:3979

By Eqs. (7.1.63), (7.1.66), and (7.1.71), the rise time, maximum overshoot, and settling time are
obtained as follows:

1  
tr ¼ π  tan1 ðωd =σÞ
ωd
1  
¼ π  tan1 ð2:3979=1:5Þ ¼ 0:888 s
2:3979
pffiffiffiffiffiffiffi2ffi
Mp ¼ eπξ=

1ξ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ exp π  0:5303= 1  0:53032 ¼ 14:0%

4 4
ts ¼ ¼ ¼ 2:667 s
ξωn 1:5
 7.1 System Response Analysis in the Time Domain 535

Example 7.1.12
For the step response of an underdamped spring–mass–damper system, m€x þ cx_ þ kx ¼ f , derive
its rise time, maximum overshoot, and settling time (1% criterion) in terms of the system
parameters (m, c, k).

Solution
According to Example 7.1.7, the natural frequency and damping ratio of the system are given by
rffiffiffiffi
k 1 pffiffiffiffiffiffi
ωn ¼ ; ξ ¼ c= mk
m 2

Also, the roots of the characteristic equation ms2 þ cs þ k ¼ 0 are

s1 1  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ σ  jωd ¼ c  j 4mk  c2
s2 2m
pffiffiffiffiffiffiffi
with j ¼ 1. This implies that
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
c 4mk  c2
σ¼ ; ωd ¼
2m 2m
pffiffiffiffiffiffi
where the conditions 2 mk > c > 0 for underdamping have been applied. It follows that

1   fπ  tan1 ðpffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4mk=c2  1Þg
1
tr ¼ π  tan ðωd =σÞ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ωd
k=m  ðc=mÞ2 =4
( )
πσ=ωd π
Mp ¼ e ¼ exp pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4mk=c2  1
4:6 9:2m
ts ¼ ¼
σ c

Three Sets of Parameters in Design

In modeling and analysis of underdamped second-order systems, there are three sets of
parameters: (i) physical parameters, such as m, c, and k; (ii) model parameters ωn and ξ; and
(iii) performance specification parameters, such as tr , Mp , and ts . As shown in Example 7.1.12,
the physical parameters and performance specification parameters are linked by the model
parameters:
536 7 System Response Analysis

  
Physical Model Performance specification
⇔ ⇔
parameters parameters: ωn ; ξ parameters : tr ; Mp ; ts

In design of an underdamped second-order system based on its step response, certain

performance specifications are first prescribed, such as
tr ≤ tro ; mL ≤ Mp ≤ mH ; ts ≤ tso (7.1.72)

where tro ; mL ; mH ; tso are specified values. The performance specification parameters are then
related to the physical parameters through the model parameters, converting the perform-
ance specifications (7.1.72) into a set of conditions on the physical parameters. With these
conditions, the values and/or ranges of the physical parameters are determined so that the
system response satisfies the performance specifications.

Example 7.1.13
YðsÞ b
For a system ¼ , with b > 0, determine the parameters a and b, such that its
RðsÞ s2 þ ð2 þ aÞs þ b
step response meets the following two performance specifications: (i) Mp ≤ 10%; and (ii) ts ¼ 3 s
with a 2% criterion. Plot the response of the system with the selected a and b, subject to a unit step
input rðtÞ ¼ 1.

Solution
The characteristic equation of the system is s2 þ ð2 þ aÞs þ b ¼ 0, which by comparison with Eq.
(7.1.34) gives 2ξωn ¼ ð2 þ aÞ; ω2n ¼ b, or

1 1
ξωn ¼ ð2 þ aÞ; ξ2 ¼ ð2 þ aÞ2
2 4b

For the condition on the settling time,

4 4
ts ¼ ¼ ¼3
ξωn 1
ð2 þ aÞ
2
which gives a ¼ 2=3. For the condition on the maximum overshoot
pffiffiffiffiffiffiffi2ffi
Mp ¼ eπξ= 1ξ
≤ 0:1

By Eq. (7.1.67), the above condition can be written as

ðln 0:1Þ2
ξ2≥ ¼ 0:3494
π2 þ ðln 0:1Þ2
 7.1 System Response Analysis in the Time Domain 537

Thus,
1 1
ξ2 ¼ ð2 þ aÞ2 ¼ ð2 þ 2=3Þ2 ≥ 0:3494
4b 4b

where a ¼ 2=3 has been used. This leads to b ≤ 5:088. Hence, the selection of the parameters:

2
a ¼ ; 0 < b ≤ 5:088
3

meets the performance specifications. For a simulation, select a ¼ 2=3 and b ¼ 4. The transfer
YðsÞ 4
function then becomes ¼ 2 . The step response is plotted by MATLAB in
RðsÞ s þ ð8=3Þs þ 4
Figure 7.1.16, showing a maximum overshoot of 6% and a settling time of 3 s.

7.1.5 Time Responses of Second-Order Systems to a Sinusoidal Input

So far, we have considered time responses subject to impulse, step, and ramp inputs. In this
section, time responses of second-order systems to a sinusoidal input are examined. A general
treatment of dynamic systems under sinusoidal excitations is presented in Section 7.3.

Undamped Systems under a Sinusoidal Input

The forced response of an undamped system (ξ = 0) under a sinusoidal input is governed by

1.2

0.8
y(t)

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3 3.5 4
t (seconds)

Figure 7.1.16 Step response of the system in Example 7.1.11

538 7 System Response Analysis

y€ þ ω2n yðtÞ ¼ ω2n r0 sin ωt

(7.1.73)
yð0Þ ¼ 0; _
yð0Þ ¼0

Here, without a loss of generality, the input rðtÞ ¼ r0 sin ωt is used. The solution of Eq.
(7.1.73), according to Section 2.6, is of the form

yðtÞ ¼ A cos ωn t þ B sin ωn t þ yp ðtÞ (7.1.74)

where yp is a particular solution, and A and B are constants that are determined by the zero
initial conditions. If the excitation frequency is not the same as the natural frequency of the
ω2 r0
system, ω ≠ ωn , yp ðtÞ ¼ 2 n 2 sin ωt, and the response is determined as
ωn  ω

ω2 r0 ω
yðtÞ ¼ 2 n 2 sin ωt  sin ωn t (7.1.75)
ωn  ω ωn

If the excitation frequency is identical to the natural frequency of the system, ω ¼ ωn , an

unbounded oscillatory response called resonant response occurs. This is seen from
1
Table 2.6.2, from which, yp ðtÞ ¼  ωn r0 t cos ωn t. Thus, with Eq. (7.1.74) and the zero
2
initial conditions, the response of the system is obtained as

1
yðtÞ ¼ ωn r0 ðsin ωn t  t cos ωn tÞ (7.1.76)
2

As can be seen, the term t cos ωn t in Eq. (7.1.76) renders the response unbounded and
oscillatory. Resonant vibrations of mechanical systems are commonly seen in engineering
applications.
For comparison, the response of a system with ωn ¼ 4 is plotted versus time in
Figure 7.1.17 in two sinusoidal input cases: (a) ω ¼ 3; and (b) ω ¼ 4. Here, a unity input
amplitude ðr0 ¼ 1Þ is assumed and Eqs. (7.1.75) and (7.1.76) are used in computation.

Damped Systems under a Sinusoidal Input

The forced response of a damped system ðξ > 0Þ under a sinusoidal input is described by

_ þ ω2n yðtÞ ¼ ω2n r0 sin ωt

y€ þ 2ξωn yðtÞ
(7.1.77)
_
yð0Þ ¼ 0; yð0Þ ¼0

From Section 2.6, it can be shown that the forced response is of the form

yðtÞ ¼ r0 H sinðωt  θÞ þ eσt ϕðtÞ (7.1.78)

 7.1 System Response Analysis in the Time Domain 539

4 60

3
40
2
20
1

y(t)
y(t)

0 0

–1
–20

–2
–40
–3

–4 –60
0 5 10 15 20 25 30 0 5 10 15 20 25 30
t t
(a) (b)

Figure 7.1.17 Response of an undamped system ðωn ¼ 4Þ to a sinusoidal excitation ðr0 ¼ 1Þ: (a) ω ¼ 3;
(b) ω ¼ 4, plotted with MATLAB

where

ω2n 1 2ξωn ω
H ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; θ ¼ tan (7.1.79)
ω2  ω2
ðω2n  ω2 Þ2 þ 4ξ 2 ω2n ω2 n

and ϕðtÞ is a known function such that eσt ϕðtÞ → 0 as t goes to infinity. Note that with
a nonzero ξ, H is finite at any frequency ω. Therefore, the system response is always bounded.
It is easy to see that a steady-state response in the sense of lim yðtÞ does not exist. However,
t→∞
another type of steady-state response can be defined by the remaining part of Eq. (7.1.78) as
t → ∞; which is given by

yss ðtÞ ¼ r0 H sinðωt  θÞ (7.1.80)

This indicates that after a long enough time, the response of a damped system becomes
sinusoidal with the same frequency as the input. This type of steady-state response is studied
further in Section 7.3.

7.1.6 Time Responses of Higher-Order Systems

We first examine third-order systems, and then consider general high-order systems. The free
response of a high-order system can be obtained by the methods mentioned in Section 7.1.1.
In this section, we focus on forced responses, especially step responses and steady-state
responses.
540 7 System Response Analysis

Third-Order Systems
For a third-order system, its transfer function is of the general form

YðsÞ b2 s2 þ b1 s þ b0
¼ GðsÞ ¼ 3 (7.1.81)
RðsÞ s þ a2 s2 þ a1 s þ a0

By partial fraction expansion (Section 2.4.4), the transfer function falls into the following two
cases.

Case 1. G(s) with Three Real Roots

In this case, the transfer function is in one of the following three forms

K1 K2 K1
GðsÞ ¼ þ þ (7.1.82a)
s þ σ1 s þ σ2 s þ σ3
B1 B2 K
GðsÞ ¼ þ þ (7.1.82b)
ðs þ σ1 Þ2 s þ σ1 s þ σ2
B1 B2 B3
GðsÞ ¼ þ þ (7.1.82c)
ðs þ σ1 Þ3 2
ðs þ σ1 Þ s þ σ1

where σ1 ;  σ2 ; and  σ3 are the transfer function poles. Without a loss of generality,
assume that σk > 0. Note that Eq. (7.1.82a) is about the system with distinct poles, and that
Eqs. (7.1.82b) and (7.1.82c) are about repeated poles. For a system of distinct poles subject to
a step input rðtÞ ¼ r0 , the system response in the s-domain, from Eq. (7.1.82a), is

β1 r0 β2 r0 β3 r0
YðsÞ ¼ þ þ (7.1.83)
T1 s þ 1 s T2 s þ σ2 s T3 s þ 1 s

where Tj ¼ 1=σj and βj ¼ Kj =σj , j = 1, 2, 3. It follows that the step response of the third-order
system is
t !
 
t ! 
t !
yðtÞ ¼ r0 β1 1e T 1
þ r0 β2 1  e T 2
þ r0 β3 1  e 3T (7.1.84)

The steady-state response of the system is then given by


K1 K2 K3
yss ¼ lim yðtÞ ¼ r0 þ þ (7.1.85)
t→∞ σ1 σ2 σ3

Systems with repeated poles can be treated similarly. The step response of the system with
repeated poles can be similarly obtained.
 7.1 System Response Analysis in the Time Domain 541

Case 2. G(s) with a Real Root and a Pair of Complex/Imaginary Poles

The system transfer function can be written as

K bs þ c
GðsÞ ¼ þ 2 (7.1.86)
s þ a s þ 2ξωn s þ ω2n

where 0 ≤ ξ < 1. Without loss of generality, we assume that a > 0. The poles of the system
pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffi
are distinct: a;  σ  jωd , with σ ¼ ξωn , ωd ¼ 1  ξ 2 ωn , and j ¼ 1. Under a step
input rðtÞ ¼ r0 , the system response in the s-domain is written as

β r0 ω2 ðBs þ CÞ r0
YðsÞ ¼ þ 2 n (7.1.87)
Ts þ 1 s s þ 2ξωn s þ ω2n s

with T ¼ 1=a, β ¼ K=a, B ¼ b=ω2n , and C ¼ c=ω2n . Thus, by Eqs. (7.1.24) and (7.1.57), the
step response of the system is
0 1

t
ωn
yðtÞ ¼ r0 β@1  e A þ r0 B pffiffiffiffiffiffiffiffiffiffiffiffiffi eσt sin ωd t
T
1  ξ2
(7.1.88)
 
σ
þ r0 C 1  eσt cos ωd t þ sin ωd t
ωd

The steady-state response of the system is yss ¼ lim yðtÞ ¼ r0 ðβ þ CÞ.

t→∞
As shown in Eqs. (7.1.84) and (7.1.88), the step response of a third-order system is a sum of
the responses of first-order and/or second-order systems.

General High-Order Systems

For an n-th-order system, its transfer function given in Eq. (7.1.6) can be written as

Xn1
ki X
n2
bi s þ ci
GðsÞ ¼ þ ; n ¼ n1 þ n2 (7.1.89)
i¼1
s þ ai i¼1 s þ 2ξ i ωi s þ ω2i
2

where partial fraction expansion has been applied. This implies that the system is
a combination (parallel connection) of n1 first-order subsystems and n2 second-order sub-
systems; see Figure 7.1.18. Due to the complexity of such a system, analytical expressions of
performance specification parameters, such as the rise time, maximum overshoot, and
settling time for a second-order system, are extremely difficult to obtain, if not impossible.
Under the circumstances, the time response of a high-order system is usually determined
numerically by software packages, such as MATLAB and Mathematica.
542 7 System Response Analysis

Figure 7.1.18 A high-order system as a combination of first- and second-order subsystems

Example 7.1.14
YðsÞ 3s þ 5
For a fourth-order system ¼ , its response to a unit step input is
RðsÞ s4 þ 2s3 þ 8s2 þ 4s þ 5
plotted in Figure 7.1.19 by MATLAB, with the following commands:
≫ num = [3 5]; den = [1 2 8 4 5];
≫ step(num, den, ‘k’)
Step Response
1.6

1.4

1.2

1
Amplitude

0.8

0.6

0.4

0.2

0
0 5 10 15 20 25 30
Time (seconds)

Figure 7.1.19 Step response of the system in Example 7.1.14

 7.1 System Response Analysis in the Time Domain 543

Im Im

P1 jω
P

−σ Re −σ Re
P2 x

a a

Figure 7.1.20 Dominant poles: (a) one real dominant pole, p ¼ σ; (b) two complex dominant poles,
p1;2 ¼ σ  jω

Dominant Poles and Model Reduction

Assume that all the poles of the transfer function GðsÞ given in Eq. (7.1.89) have negative
parts. (The system is stable; see Section 7.2.2.) The pole that is closest to the imaginary axis is
called the dominant pole of the system. There are two cases of dominant poles: (i) one real
pole p ¼ σ, and (ii) two complex conjugate poles p1;2 ¼ σ  jω; see Figure 7.1.20, where
σ > 0, and a is the shortest distance between all other poles and the imaginary axis.
As a rule of thumb, if a ≥ 10σ, the n-th-order system described by Eq. (7.1.89) can be
approximated as a first-order system

K K
GðsÞ ≈ ¼ (7.1.90a)
sp sþσ

or as a second-order system
K1 K2 As þ B
GðsÞ ≈ þ ¼ (7.1.90b)
s  p1 s  p2 ðs þ σÞ2 þ ω2

where K, K1 , and K2 are the transfer function residues determined by partial fraction
expansion.
Even if the condition of a ≥ 10σ is not met, the concept of dominant poles is still useful in
design of feedback-control systems; see Section 8.6, where the root locus method is
presented.

Example 7.1.15
Consider a third-order system with the transfer function

4 1
GðsÞ ¼  ; a>1 (a)
s2 þ 1:2s þ 4 s=a þ 1
which has poles 0:6  j1:9079 and a. The dominant poles of the system are 0:6  j1:9079,
with σ ¼ 0:6. For a ≥ 10σ ¼ 6, the system can be approximated as a second-order system
544 7 System Response Analysis

4
GðsÞ ≈ (b)
s2 þ 1:2s þ 4

Note that the third-order transfer function in Eq. (a) is reduced to the second-order transfer
function in Eq. (b) when a → ∞. This means that the pole a has an insignificant effect on the
system response if it is located far away from the imaginary axis.
To validate the approximate model, consider the response of the third-order system subject to
a step input rðtÞ ¼ 1, with a ¼ 0:6; 1:2; 6; 12, and ∞. Here, a ¼ ∞ represents the second-order
system given in Eq. (b). The step response of the system with different values of a is plotted in
Figure 7.1.21. As seen from the figure, the step response curves with a = 6 and 12 are close to the
solid curve, which is for the second-order system ða ¼ ∞Þ. On the other hand, with a = 0.6 or 1.2,
the step response curve has much deviation from the solid curve. Therefore, the approximate
model given in Eq. (b) gives acceptable response results for a ≥ 6.

Step Response
1.4

1.2

0.8
y(t)

0.6
a = 0.6
a = 1.2
0.4
a=6
a = 12
0.2
a=∞

0
0 1 2 3 4 5 6 7 8 9 10
t

Figure 7.1.21 Step response of the third-order system in Example 7.1.15: a ¼ 0:6; 1:2; 6; 12,
and ∞

7.2 Stability Analysis

In analysis and design of dynamic systems, especially feedback control systems, stability is
essentially important. The stability of a system is related to the boundedness of its response to
initial and external disturbances. An unstable system generally has an unbounded response.
 7.2 Stability Analysis 545

Such an unbounded response makes it difficult to achieve desired system performance, can
cause malfunction and failure of the system, and in a worst-case scenario even leads to
catastrophic consequences, including property damage and loss of human life. Therefore, in
design and operation of dynamic systems, assurance of stability is a must.

7.2.1 Stability Definitions

In this text, we shall only consider stability analysis of linear time-invariant systems.
According to Section 7.1, the total response of a linear time-invariant system is the sum of
its free response and forced response; namely,

yðtÞ ¼ yI ðtÞ þ yF ðtÞ

Stability can be defined by either the free response or the forced response.
First, consider stability definitions by forced responses. A system is stable if its response is
bounded in magnitude to every bounded input. A system is unstable if there is a bounded
input that yields an unbounded output. These are stability definitions in the sense of bounded
input–bounded output (BIBO).
For a bounded input, say jrðtÞj ≤ M < ∞, Eq. (7.1.11) gives
ðt ðt
jyF ðtÞj ≤ jgðt  τÞjjrðτÞjdτ ≤ M jgðt  τÞjdτ
0 0
where gðtÞ is the impulse response function of the system. This implies that the system is
Ð∞
BIBO stable if jgðtÞjdτ is finite. As an example, consider an underdamped second-order
0
system with the impulse response
ωn
gðtÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi eσt sin ωd t
1  ξ2

which is given in Eq. (7.1.52b). Because

ð∞ ð∞
ωn 1
jgðtÞjdτ ≤ pffiffiffiffiffiffiffiffiffiffiffiffiffi eσt dτ ¼ < ∞; for 0 < ξ < 1
1  ξ2 1  ξ2
0 0

the system is BIBO stable.

Next, consider stability definitions with free responses. A linear time-invariant system can
have the following three states of stability:

(i) A system is asymptotically stable if its free response approaches zero as time goes to
infinity.
546 7 System Response Analysis

B
A C

Figure 7.2.1 Illustration of stability states on a ball

(ii) A system is unstable if its free response grows unbounded as time goes to infinity.
(iii) A system is marginally stable if its free response does not vanish and remains bounded as
time goes to infinity.

It can be shown that an asymptotically stable system is also a BIBO stable system. Thus, an
asymptotically stable system is simply called a stable system.
It should be pointed out that the above stability definitions about a linear time-invariant
system can also be given by its impulse response. In other words, equivalent stability
conditions can be stated with “free response” replaced by “impulse response.”
Physically, initial disturbances, such as the initial displacement and velocity of
a mechanical system, describe a finite energy input at the initial time. A bounded input, on
the other hand, can input infinite energy to a system in the entire time region, 0 < t < ∞.
Examples include a step input and a sinusoidal input of properly tuned frequency. So,
a bounded free response does not necessarily guarantee a bounded forced response. This
implies that a marginally stable system may be BIBO unstable. For instance, an undamped
spring–mass system has bounded free response, but it experiences unbounded resonant
vibration under a sinusoidal input with the frequency identical to the natural frequency of
the system, as shown in Eq. (7.1.75).
The above three stability states with respect to free response can be illustrated in
Figure 7.2.1, where a ball, under gravity, sits at different positions of equilibrium: point A,
which is the bottom of a valley; point B, which is the peak of a mountain; and point C on
a horizontal surface. Assume that the ball experiences friction and damping in motion. If the
ball is slightly disturbed at point A, it moves back and forth in the valley, and eventually
returns to the original equilibrium position. Point A and the response are said to be stable. If
the ball is disturbed at point B, it falls off the peak, and never comes back to the original
equilibrium point. Point B and the response are said to be unstable. If the ball is disturbed at
point C, it travels a finite distance, and eventually settles at a new equilibrium position on the
horizontal plane. Point C and the response are said to be marginally stable.

7.2.2 Stability Conditions in Terms of Pole Locations

In the subsequent stability analysis, the three stability definitions based on free response are
used. For an n-th-order system, its characteristic equation is

AðsÞ ¼ an sn þ an1 sn1 þ . . . þ a1 s þ a0 ¼ 0 (7.2.1)

 7.2 Stability Analysis 547

with the roots being the poles of the transfer function given in Eq. (7.1.6). According to Eqs.
(7.1.9) and (7.1.10), the system free response can be written as

yI ðtÞ ¼ k1 ep1 t þ k2 ep2 t þ . . . þ kn epn t (7.2.2)

if the system has a distinct pole p1 ; p2 ; . . . ; pn , or

 
yI ðtÞ ¼ b1 tl1 þ b2 tl2 þ . . . þ bl1 t þ bl ep0 t þ k1 ep1 t þ k2 ep2 t þ . . . þ km epm t (7.2.3)

if the system has l identical poles p0 and m distinct poles p1 ; p2 ; . . . ; pl , with l þ m ¼ n and
p0 ≠ pk . Here, the constants ki and bj are determined by the initial conditions, as given by Eq.
(7.1.2). A system with more repeated roots can be treated similarly.
According to Eqs. (7.2.2) and (7.2.3), the free response of a system vanishes as t approaches
infinity if all the poles of the system have negative real parts. This means that the poles of a stable
system must lie in the left half-plane, excluding the imaginary axis. For the free response to be
bounded, all the poles must have nonpositive real parts and there are no repeated poles on the
imagery axis (jω-axis), including the origin of the complex plane. Furthermore, the free response
is unbounded if at least one pole has a positive real part or if there are repeated poles on the
imaginary axis. The correlation between system stability and pole locations in the complex plane
are illustrated in Figure 7.2.2, where the crosses ðÞ indicate a pole location.
Based on the above analysis and discussion, the following three stability conditions in
terms of pole locations are established.
Condition 1 for stability. The necessary and sufficient condition for a system to be stable is
that all the poles of the system must have negative real parts. In other words, a system is

Im Im

Re Re

Stable Marginally stable

Im Im

Re Re

Unstable Unstable

Figure 7.2.2 Stability and poles: – pole location in the complex plane
548 7 System Response Analysis

stable if and only if all the poles of the system are in the left half-plane, excluding the
imaginary axis.
Condition 2 for instability. A system is unstable if one of the following two conditions is met:
(i) at least one pole is in the right half-plane, with positive real part; or
(ii) there are repeated poles on the imaginary axis, including the origin of the complex plane.
Condition 3 for marginal stability. A system is marginally stable if some simple (distinct) poles
are on the imaginary axis (including the origin, s = 0) and if all other poles are in the open
left half-plane.

Example 7.2.1
Consider the characteristic equations of the following four systems
System 1: ðs þ 1Þðs2 þ 2s þ 5Þ ¼ 0, with poles 1;  1  2j
System 2: ðs þ 1Þðs2  2s þ 5Þ ¼ 0, with poles 1; 1  2j
System 3: ðs þ 1Þðs2 þ 4Þ ¼ 0, with poles 1;  2j
System 4: s2 ðs þ 1Þ ¼ 0, with poles 0; 0;  1
From the stability conditions, System 1 is stable, Systems 2 and 4 are unstable, and System 3 is
marginally stable.

Example 7.2.2
For the four mechanical systems in Figure 7.2.3, the transfer functions are as follows:
X1 ðsÞ 1 1  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
System 1: G1 ðsÞ ¼ ¼ 2 ; with poles c  c2  4mk
FðsÞ ms þ cs þ k 2m

Figure 7.2.3 Four mechanical systems in Example 7.2.2

 7.2 Stability Analysis 549

X2 ðsÞ 1 pffiffiffiffiffiffiffiffiffi
System 2: G2 ðsÞ ¼ ¼ 2 ; with poles  j k=m
FðsÞ ms þ k
X3 ðsÞ 1
System 3: G3 ðsÞ ¼ ¼ 2 ; with poles 0;  c=m
FðsÞ ms þ cs
X4 ðsÞ 1
System 4: G4 ðsÞ ¼ ¼ 2 ; with poles 0; 0
FðsÞ ms
From the stability conditions, System 1 is stable, Systems 2 and 3 are marginally stable, and
System 4 is unstable.
To show the stability states of these systems, consider m = 1, c = 2, and k = 16. Also, consider an
impulsive force f ðtÞ ¼ δðtÞ. Here, the nondimensional system parameters are used. The impulse
response, which can also be used to define stability, is plotted for the four systems by MATLAB in
Figure 7.2.4.

0.2 0.25

0.2
0.15
0.15

0.1
0.1
0.05
x1(t)

x2(t)

0.05 0

−0.05
0
−0.1

−0.15
−0.05
−0.2

−0.1 −0.25
0 1 2 3 4 5 6 0 1 2 3 4 5 6
t (seconds) t (seconds)
(a) (b)
0.5 50

0.45 45

0.4 40

0.35 35

0.3 30
x3 (t)

x4(t)

0.25 25

0.2 20

0.15 15

0.1 10

0.05 5

0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 50
t (seconds) t (seconds)
(c) (d)

Figure 7.2.4 Impulse response of the mechanical systems: (a) System 1; (b) System 2; (c) System 3;
and (d) System 4
550 7 System Response Analysis

Estimation of the Boundedness of Forced Response

The three stability conditions can be used to check the boundedness of a forced response
subject to a specific input. As matter of fact, Eqs. (7.2.2) and (7.2.3) are the results of partial
fraction expansion of IðsÞ=AðsÞ, where IðsÞ is a polynomial of s given in Eq. (7.1.5a), with the
order of less than n. Let the Laplace transform of the input rðtÞ be RðsÞ ¼ BR ðsÞ=AR ðsÞ, where
AR ðsÞ and BR ðsÞ are polynomials of s. The s-domain forced response by Eq. (7.1.5b) is

NðsÞ
YF ðsÞ ¼ GðsÞRðsÞ ¼
DðsÞ

with DðsÞ ¼ AðsÞAR ðsÞ and NðsÞ ¼ BðsÞBR ðsÞ. If the order of DðsÞ is larger than that of NðsÞ,
Eqs. (7.2.2) and (7.2.3) are directly applicable to NðsÞ=DðsÞ and, as such, the stability
conditions can be used to check the boundedness of the forced response yF ðtÞ ¼
L1 fNðsÞ=DðsÞg.

Example 7.2.3 pffiffiffiffiffiffiffiffiffi

Let System 2 in Example 7.2.2 be subject to a sinusoidal input f ðtÞ ¼ f0 sin ωn t, with ωn ¼ k=m.
The s-domain system response is

1 f0 ωn
X2 ðsÞ ¼ G2 ðsÞFðsÞ ¼ :
m ðs2 þ ωn Þ ðs þ ω2n Þ
2 2

pffiffiffiffiffiffiffi
which has repeated poles on the imaginary axis: jωn ; jωn ; jωn ; jωn , j ¼ 1. From
Condition 2, the time response x2 ðtÞ is unbounded.
Also, let System 3 in Example 7.2.2 be subject to a step input f ðtÞ ¼ f0 . The s-domain response is

1 f0
X3 ðsÞ ¼ G2 ðsÞFðsÞ ¼
sðms þ cÞ s

which has repeated poles on the imaginary axis: 0, 0. From Condition 2, the time response x3 ðtÞ is
unbounded.
Because the sinusoidal and step inputs are bounded, Systems 2 and 3, which are marginally
stable, are unstable in the sense of BIBO.

7.2.3 The Routh–Hurwitz Stability Criterion

The stability conditions in the previous section require the knowledge of system poles to
determine the stability status. In this section we study a method that does not require solving
 7.2 Stability Analysis 551

the characteristic equation (7.2.1) for the system poles. This method, namely, the Routh–
Hurwitz stability criterion, has two special features: first, it does not require the knowledge
of system poles to gain stability information; second, it can deal with a characteristic
equation with coefficients consisting of unknown parameters, and it can obtain the ranges
of these parameters for stability. Therefore, the Routh–Hurwitz stability criterion is
a useful method in analysis and design of dynamic systems, especially feedback-control
systems.
Let the characteristic equation of a system be written as

AðsÞ ¼ sn þ an1 sn1 þ . . . a1 s þ a0 ¼ 0 (7.2.4)

where, for convenience of analysis, the polynomial AðsÞ from Eq. (7.2.1) has been
scaled such that an ¼ 1. With a data table called a Routh array, which is constructed
by using the coefficients of AðsÞ, the Routh–Hurwitz criterion tells how many poles
are in the left half-plane, in the right half-plane, and on the imaginary axis.
We first present a necessary condition for stability, then show how to construct a Routh
array, and finally introduce the Routh–Hurwitz stability criterion, which is a necessary and
sufficient condition for stability.

Necessary Condition for Stability

A necessary condition for all the roots of the characteristic equation (7.2.4) to have negative
real parts is that all the coefficients of AðsÞ are positive; that is,

ai > 0; i ¼ 0; 1; 2; . . . ; n  1 (7.2.5)

This means that for a stable system, all the coefficients of its characteristic equation,
Eq. (7.2.4), must be positive, and none of the coefficients vanish.
The proof of the necessary condition is straightforward. Let a stable system have l real
pffiffiffiffiffiffiffi
roots ðσi ; i ¼ 1; 2; . . . ; lÞ and 2m complex roots ðαk  jβk ; k ¼ 1; 2; . . . ; m; j ¼ 1Þ,
with l þ 2m ¼ n. Because the system is stable, σi > 0 for all i and αk > 0 for all k. The
characteristic equation becomes

l m  
AðsÞ ¼ ∏ ðs þ σi Þ ∏ ðs þ αk Þ2 þ β2k ¼ 0
i¼1 k¼1

Because σi , αk and β2k are positive numbers for all i and k, a comparison of the previous
equation with Eq. (7.2.4) reveals that all the coefficients of AðsÞ are positive.
The necessary condition can be used as means of initial screening for instability. In
addition, one can make use of the following two observations.
552 7 System Response Analysis

(1) If some s-power terms of AðsÞ disappear and if the remaining terms all have positive
coefficients, the system is either marginally stable or unstable.
(2) If at least one coefficient of AðsÞ is negative, the system is unstable.

However, even if the condition (7.2.5) is met, the stability of the system still cannot be
concluded. Further effort, for example analysis by the Routh–Hurwitz criterion, is required
to detect the system stability.

Example 7.2.4
Consider the characteristic equations of the following three systems
System 1: s3 þ 2s2 þ 5s þ 6 ¼ 0
System 2: s3 þ 2s2 þ 6 ¼ 0
System 3: s4 þ 3s3  2s2 þ 6s þ 20 ¼ 0
System 1 satisfies the necessary condition, and as such its stability needs further investigation.
System 2, with a missing s1 -term, is either marginally stable or unstable. System 3, with the
coefficient of the s2 -term being negative, is unstable. For validation purposes, the poles of these
systems are computed by the MATLAB function roots and the corresponding stability states are
concluded as follows:
System 1: 0:2836  2:0266i, 1:4329, stable
System 2: 0:3888  1:4174i, 2:7777, unstable
System 3: 1:0832  1:5365i, 1:5761;  3:5904, unstable

Routh Array
The application of the Routh–Hurwitz criterion takes two steps: (i) to generate a Routh
array; and (ii) to examine the array to detect the system stability. We now show how to
generate such an array.
The Routh array for the characteristic equation (7.2.4) is a table of n + 1 rows, as shown
below:


sn

1 an2 an4 an6 ...
sn1

an1 an3 an5 an7 ...
sn2

b1 b2 b3 b4 ...
sn3

c1 c2 c3 c4 ...
..
. ..
.
.. .


s2

d1 d2
s1

e1
s0
f 1
 7.2 Stability Analysis 553

The rows are labeled by the powers of s, from sn to s0 . The first row (the sn row) starts with the
coefficient 1 of the sn term (the highest s-power term) of AðsÞ, and lists every other coefficient
of the polynomial. The second row (the sn1 row) starts with the coefficient an1 of the sn1
term of AðsÞ, and lists every other coefficient of the polynomial. The remaining rows (from
the sn2 row to the s0 row) are formed as follows:







1

1 an2


1

1 an4


1

1 an6



b1 ¼

; b2 ¼

; b3 ¼

; . . .
an1
an1 an3
an1
an1 an5
an1
an1 an7








1

an1 an3

1

an1 an5

1

an an7


c1 ¼

; c ¼

; c ¼

; . . .
b1
b1 b2
2 b1
b1 b3
3 b1
b1 b4

......



1

d1

d2

f1 ¼

¼ d2
e1
e1 0

As can be seen, the calculation for each element involves a two-by-two determinant and
a common factor.
The formation of the rows beneath the sn1 row follows several rules, as stated below:

Rule 1. The current row is generated by using the elements in the previous two rows, and the
creation of new elements stops when all the elements in the previous two rows have been
exhausted.
Rule 2. The elements in a row share the same factor, which is −1 divided by the first element in
the previous row. For instance, in the sn3 row, the elements c1 ; c2 ; c3 ; . . . have a common
factor 1=b1 .
Rule 3. In the two-by-two determinant for an element, the left-hand column is always the first
column of the previous two rows, and the right-hand column is a column selected from the
previous two rows, from the second column to the last one. For example, in the sn3 row,

a
the left-hand column for each determinant is n1 , and the right-hand columns for
b1
  
a a a
elements c1 ; c2 ; c3 ; . . . are n3 , n5 , n7 ; . . . , respectively.
b2 b3 b4
Rule 4. A zero is added in the determinant for the last element in a row if the previous two
rows have unequal numbers of elements. See the calculation of f1 for instance.
Rule 5. For convenience in calculations, any row can be multiplied or divided by a positive
number.
According to Rule 3, the number of elements in a row decreases as the s-power label index
decreases. In particular, the s2 row always has two elements, and the s1 and s0 rows each only
have one element. Because of this, the single element f1 of the last row is equal to d2 , which is
the second element in the s2 row.
554 7 System Response Analysis

Example 7.2.5
Given the characteristic equation AðsÞ ¼ s4 þ 20s3 þ 8s2 þ 200s þ 14 ¼ 0, the corresponding
Routh array is created as follows:

In construction of the previous array, Rule 5 is applied to the s3 row, Rule 4 is applied to the s2
row, and f1 ¼ d2 is directly used for the s0 row without calculation.

Routh–Hurwitz Criterion
Once the Routh array for a system is generated, the stability of the system can be examined
by the Routh–Hurwitz criterion as follows:
The number of the roots of A(s) with positive real parts is equal to the number of sign changes in the first
column (the far-left column) of the Routh array.

For instance, in Example 7.2.5, the first column of the Routh array has two sign changes: one
from 1 of the s3 row to −2 of the s2 row, and another from −2 of the s2 row to 17 of the s1 row.
Therefore, the system is unstable with two poles in the right half-plane. To validate this,
using the MATLAB function roots gives the roots of the characteristic equation as:
−0.0702, −20.0954, and 0:0828  3:15i, showing two poles with positive real parts.
According to the Routh–Hurwitz criterion, a system is stable if there is no sign change in
the first column of the Routh array. Therefore, a necessary and sufficient condition for
a system to be stable is that all the elements in the first column of the Routh array are positive:

an1 > 0; b1 > 0; c1 > 0; . . . ; d1 > 0; e1 > 0; f1 > 0 (7.2.6)

There are some cases in which the first element of a row is zero or all the elements in a row
are zeros. Should this happen, the system in consideration is not stable. Under the circum-
stances, the construction of such a Routh array can proceed with some methods, as shown in
the references at the end of this chapter.
In summary, the application of the Routh–Hurwitz criterion in stability analysis takes the
following three steps:
 7.2 Stability Analysis 555

(i) Use the necessary condition (7.2.5) to screen for instability.

(ii) Construct a Routh array.
(iii) Apply the Routh–Hurwitz criterion to conclude the stability status.

Furthermore, in design of a stable system, the necessary and sufficient condition (7.2.6) must
be satisfied.

Example 7.2.6
Consider a system with the characteristic equation

s3 þ ða  6Þs2 þ 5s þ b ¼ 0
Determine the ranges of parameters a and b for the system to be stable.
Solution
First, check the necessary condition (7.2.5), which indicates that for the system to be stable, the
parameters should satisfy the conditions a > 6 and b > 0. Second, construct the Routh array as
follows:


s3

1 5
s2

a  6 b
s1

Δ
s0
b

with



1

1 5

5a  b  30
Δ¼ ¼
a  6
a  6 b
a6

According to the Routh–Hurwitz criterion (7.2.6), the system is stable if and only if

a  6 > 0; Δ > 0; b > 0 (a)

which can be further written as

a > 6; 0 < b < 5a  30 (b)

As can be seen, the conditions (a) include the necessary conditions.

Stability Conditions of Systems of Orders Two to Four

The Routh–Hurwitz criterion is applied to second-, third-, and fourth-order systems, and the
stability conditions are listed in Table 7.2.1. The derivation of the stability conditions is given
below.
556 7 System Response Analysis

Table 7.2.1. Stability conditions for systems of order two to four

Characteristic equation Stability conditions

Second-order s2 þ a1 s þ a0 ¼ 0 a1 > 0; a0 > 0

system
a0
Third-order s3 þ a2 s2 þ a1 s þ a0 ¼ 0 a0 > 0; a1 > > 0; a2 > 0
a2
system
a0 a3
Fourth-order s4 þ a3 s3 þ a2 s2 þ a1 s þ a0 ¼ 0 a0 > 0; a1 > >0
a2  a1 =a3
system
a1
a2 > > 0; a3 > 0
a3

For a second-order system, the Routh array is

s2

1 a0
s1

a1
s0
a0

From the Routh–Hurwitz criterion, the system is stable if and only if a1 > 0 and a0 > 0. Note
that for a second-order system the necessary and sufficient conditions are the same.
For a third-order system, the Routh array is


s3

1 a1
s2

a2 a0
s1

Δ
s0
a0

with



1

1 a1

a1 a2  a0
Δ¼ ¼
a2
a2 a0
a2

From the Routh–Hurwitz criterion, the stability conditions for the system are
a2 > 0; Δ > 0; a0 > 0
which, after algebraic manipulations, give
a0 > 0; a2 > 0; a1 > a0 =a2
Here, the condition a1 > 0 is implied by a1 > a0 =a2 because a0 > 0 and a2 > 0.
For a fourth-order system, the Routh array is


s4

1 a2 a0
s3

a3 a1
s2

b1 b2
s1

c1
s0
b2
 7.2 Stability Analysis 557

with
a2 a3  a1 a1 b1  a3 b2
b1 ¼ ; b2 ¼ a0 ; c1 ¼
a3 b1

From the Routh–Hurwitz criterion, the stability conditions for the system are
a3 > 0; b1 > 0; c1 > 0; b2 > 0
which, after algebraic manipulations, lead to
a0 a3
a0 > 0; a1 > ; a2 > a1 =a3 ; a3 > 0
a2  a1 =a3
The conditions a1 > 0 and a2 > 0 are imbedded in the above conditions.

Example 7.2.7
In Figure 7.2.5, a second-order plant is under proportional and integral (PI) control (see
Section 8.3), where kP and kI are gain constants. Determine the ranges of the gain constants for the
stability of the closed-loop system.

Solution
The characteristic equation of the closed-loop system is

1 1
1 þ kP þ k I ¼0
s sðs þ 5Þ

or
s3 þ 5s2 þ kP s þ kI ¼ 0

This is a third-order system. From Table 7.2.1, the stability conditions of the closed-loop system
are given by

1
kP > kI ; kI > 0
5

Figure 7.2.5 PI control of a second-order plant

558 7 System Response Analysis

Besides the Routh–Hurwitz stability criterion, the root locus method can be applied for
stability analysis. This method, which is a graphic technique, is especially useful in the design
of feedback control systems; see Sections 8.5 and 8.6.

7.3 System Response Analysis in the Frequency Domain

Dynamic systems subject to sinusoidal or harmonic excitations are seen in a variety of
engineering applications, including rotating machinery with unbalanced mass and electric
circuits with alternating current (AC). The term frequency response arises in the description
of the steady-state behavior of a dynamic system to a harmonic input in a range of input
frequencies. For a stable linear time-invariant system, its steady-state response to
a sinusoidal input is also sinusoidal, with the same frequency, as illustrated in Figure 7.3.1,
where the output differs from the input in both amplitude y0 and phase ϕ, which are functions
of the input frequency ω. As shown in Section 7.3.1, the frequency-dependent amplitude and
phase can be determined from the system transfer function GðsÞ with the complex parameter
pffiffiffiffiffiffiffi
s replaced by jω, j ¼ 1. The GðjωÞ is called the frequency response or frequency response
function of the system.
Note that the steady-state response mentioned herein is a sinusoidal function of time, and
it is not the same as yss ¼ lim yðtÞ, which is a constant as defined in Section 7.1.1. These two
t→∞
kinds of steady-state responses have different implications.
Although frequency response was initially used to determine the steady-state response of
stable systems, the concept of frequency response is applicable to general systems, including
unstable systems, of which a steady-state response may not exist at all. Indeed, a set of
frequency-domain tools have been developed for system modeling, stability analysis, and
feedback control design. Furthermore, frequency-domain analysis has been extended to
linear time-varying systems and even nonlinear systems.
Methods based on frequency responses have two advantages in practice. First, frequency
response designs can deliver good results for system models with uncertainties. Second,
frequency responses can be conveniently integrated with experimental data in the modeling
and feedback control of complex dynamic systems. Because of these advantages, frequency-
domain methods are widely used in industry.

r (t ) = r0 sinωt Linear
yss(t ) = y0(ω) sin(ωt +φ (ω))
system
r (t ) yss(t )

r0 y0
t t

Figure 7.3.1 Steady-state response of a linear time-invariant system

 7.3 System Response Analysis in the Frequency Domain 559

7.3.1 General Concepts

Consider a stable system subject to a sine input rðtÞ ¼ r0 sin ωt. The s-domain response of the
system, by partial fraction expansion, can be expressed as

r0 ω b1 b2 X
n
ki
YðsÞ ¼ GðsÞ ¼ þ þ (7.3.1)
s þω
2 2 s  jω s þ jω i¼1 s  pi

where GðsÞ is the system transfer function as given in Eq. (7.1.6), and pi are the poles of
the transfer function. Here, without loss of generality, distinct poles have been
assumed. Also note that b1 and b2 are a pair of complex conjugates, b2 ¼ b 1 . The time
response of the system then is obtained by inverse Laplace transform as follows:

X
n   Xn
yðtÞ ¼ b1 ejωt þ b2 ejωt þ ki epi t ¼ 2Re b1 ejωt þ ki epi t (7.3.2)
i¼1 i¼1

The residue b1 is computed by

r0 ω j
b1 ¼ lim GðsÞ ¼  r0 GðjωÞ (7.3.3)
s → jω s þ jω 2

Write
GðjωÞ ¼ A ejϕ (7.3.4)

with
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Magnitude: A ¼ jGðjωÞj ¼ ½ReðGðjωÞÞ2 þ ½ImðGðjωÞÞ2
ImðGðjωÞÞ (7.3.5)
Phase: ϕ ¼ ∠GðjωÞ ¼ tan1
ReðGðjωÞÞ

which are functions of the excitation frequency ω. Because the system is stable, the epi t related
terms in Eq. (7.3.2) vanish as time t approaches infinity. Consequently, the remaining part of
yðtÞ, which is called the steady-state response (output) and denoted by yss ðtÞ, is given by
  
 jωt  j
yss ðtÞ ¼ 2 Re b1 e ¼ 2 Re  r0 Aejϕ ejωt ¼ r0 A Re jejðωtþϕÞ
2

It follows that the steady-state response of the system under the sinusoidal input is
yss ðtÞ ¼ r0 A sinðωt þ ϕÞ (7.3.6)
where the magnitude A and phase ϕ are given by Eq. (7.3.5).
Similarly, for a more general sinusoidal input rðtÞ ¼ r0 sinðωt þ θ0 Þ, it can be shown that
the steady-state response of the system is given by
560 7 System Response Analysis

yss ðtÞ ¼ r0 A sinðωt þ θ0 þ ϕÞ (7.3.7)

Equivalently, for a sinusoidal input rðtÞ ¼ r0 cosðωt þ θ0 Þ, it can be shown that the steady-
state response of the system is

yss ðtÞ ¼ r0 A cosðωt þ θ0 þ ϕÞ (7.3.8)

Thus, for a system with the transfer function G(s), its steady-state response, which is given by
either Eq. (7.3.7) or Eq. (7.3.8), is fully determined by its frequency response (also called the
frequency response function)

GðjωÞ ¼ GðsÞj s ¼ jω (7.3.9)

Although the frequency response was originally derived from the steady-state response of
stable systems, the definition (7.3.9) has been extended to general systems, including unstable
systems, for other purposes. For an unstable system, of which a steady-state response to
a sinusoidal input generally does not exist, its frequency response is still of the form of GðjωÞ.
In general, frequency responses have the following three utilities: (a) the determination of
steady-state response for stable systems; (b) stability analysis; and (c) feedback control
design. In this chapter, we focus on the first utility, namely, the determination of steady-
state response. For the second and third utilities, refer to standard control texts, some of
which are listed at the end of the chapter.

Example 7.3.1
For the system

2s  1
GðsÞ ¼
s2 þ 2s þ 5

which is stable according to Table 7.2.1, plot the magnitude and phase of the system frequency
response against the frequency parameter ω, and determine the steady-state response of the
system subject to a cosine input rðtÞ ¼ 3 cosð4tÞ.
Solution
The frequency response of the system, by the definition given in Eq. (7.3.9), is

2jω  1
GðjωÞ ¼
5  ω2 þ 2jω

for which the magnitude and phase are

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4ω2 þ 1
A ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ð5  ω2 Þ2 þ 4ω2 (a)
 
1 2ω 2ω
ϕ ¼ ∠ð2jω  1Þ  ∠ð5  ω þ 2jωÞ ¼ tan
2
 tan
1 5  ω2
 7.3 System Response Analysis in the Frequency Domain 561

The magnitude and phase versus the frequency parameter are plotted in Figure 7.3.2.
For the given sinusoidal input, r0 ¼ 3 and ω = 4, which by Eq. (a) yields
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffi
4  42 þ 1 13
A ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi¼ ¼ 0:5927
2 37
5  42 þ 4  42
  (b)
24 24
ϕ ¼ tan1  tan
1 5  42
¼ π  tan ð8Þ  π  tan1 ð8=11Þg ¼ 0:818 rad
1

According to Eq. (7.3.8), the steady-state response of the system is

yss ðtÞ ¼ 1:778 cosð4t  0:818Þ (c)

1.2 200

1 150

0.8 100
Phase, φ (degrees)
Magnitude, A

0.6 50

0.4 0

0.2 −50

0 −100
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Frequency, ω Frequency, ω
(a) (b)

Figure 7.3.2 Frequency response plots of the system in Example 7.3.1: (a) magnitude; (b) phase

7.3.2 Frequency Response Plots

In analysis and design of dynamic systems, frequency response is usually displayed in plots,
with frequency ω as an independent variable. There are three types of commonly used
frequency response plots:

• plots of M and ϕ against ω from Eq. (7.3.5), as shown in Figure 7.3.2

• Bode diagrams
• Nyquist plots
562 7 System Response Analysis

As mentioned previously, these frequency response plots are applicable to both stable and
unstable systems. In this section, Bode and Nyquist plots are introduced. The utility of these
plots in control system design is covered in standard control textbooks.

Bode Diagrams
The Bode diagram (or Bode plot) of a frequency response function GðjωÞ consists of the
logarithmic plots of its magnitude and phase, which are defined as follows:

M ¼ 20 log10 j GðjωÞ j dB
(7.3.10)
ϕ ¼ ∠GðjωÞ

Here, the unit for M is decibel (dB). In a Bode diagram, the horizontal axis is about the
frequency parameter ω in terms of the logarithm to the base 10, namely log10 ω. On the
horizontal axis, an interval between two frequencies with a ratio of 10 is called a decade.
Bode diagrams can be easily generated by MATLAB. For instance, the Bode diagram for
the system

2s  1
GðsÞ ¼
s2 þ 2s þ 5

which is considered in Example 7.3.1, is plotted in Figure 7.3.3 by the MATLAB commands:
≫ num = [2 -1]; den = [1 2 5];
≫ bode(num, den)
≫ grid

Here, bode is a MATLAB function for plotting Bode diagrams; see Table B9 in Appendix B.
Bode diagrams have the following additive feature. For a frequency response as a product
of r factors

GðjωÞ ¼ G1 ðjωÞ G2 ðjωÞ . . . Gr ðjωÞ (7.3.11)

its Bode diagram can be written as

M ¼ M1 þ M2 þ . . . þ Mr
(7.3.12)
ϕ ¼ ϕ1 þ ϕ2 þ . . . þ ϕr

where Mk and ϕk are the magnitude and phase of the k-th factor, namely,

Mk ¼ 20 log10 j Gk ðjωÞ j dB
(7.3.13)
ϕk ¼ ∠Gk ðjωÞ

The additive feature makes the use of Bode diagrams convenient in design of feedback
control systems.
 7.3 System Response Analysis in the Frequency Domain 563

Bode Diagram
0

−10
Magnitude (dB)

−20

−30

−40

180
Phase (deg)

90

−90
10 −2 10 −1 100 101 102
Frequency (rad/s)

Figure 7.3.3 Bode diagram of the system in Example 7.3.1

Nyquist Plots
The Nyquist plot of a frequency response is a plot in the complex plane that is obtained from

GðjωÞ ¼ RðωÞ þ jX ðωÞ; for  ∞ < ω < ∞ (7.3.14)

where RðωÞ and X ðωÞ are the real and imaginary parts of the frequency response GðjωÞ, for
any given value of ω. Nyquist plots are widely used in stability analysis and the design of
feedback control systems.
Nyquist plots can also be generated by MATLAB. For instance, for the transfer function
considered in Example 7.3.1, its Nyquist plot is plotted in Figure 7.3.4 by the commands:
≫ syst = tf([2 -1], [1 2 5])
≫ nyquist(syst)

Here, nyquist is a MATLAB function for plotting Nyquist plots; see Table B9 in
Appendix B.
564 7 System Response Analysis

Nyquist Diagram
0.8

0.6

0.4

Imaginary Axis 0.2

−0.2

−0.4

−0.6

−0.8
−1 −0.5 0 0.5 1 1.5
Real Axis

Figure 7.3.4 Nyquist plot of the system in Example 7.3.1

7.3.3 Frequency Response of First-Order Systems

For a standard first-order system, as described by Eq. (7.1.18), its frequency response is


1

1
GðjωÞ ¼
¼ (7.3.15)
Ts þ 1
1 þ jωT
s¼jω

with the magnitude and phase given by

1
A ¼ jGðjωÞj ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 þ ω2 T 2 (7.3.16)
1
ϕ ¼ ∠GðjωÞ ¼ tan ðωTÞ

The magnitude of the Bode diagram is


1  
M ¼ 20log10 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ 10log10 1 þ ω2 T 2 (7.3.17)
1 þ ω2 T 2

The phase of the Bode diagram is the same as that given in Eq. (7.3.16).
The Bode diagram of the system is shown in Figure 7.3.5, where the MATLAB function
bode has been used. As can be seen from the figure and Eq. (7.3.17), the magnitude curve has
two asymptotes:
 7.3 System Response Analysis in the Frequency Domain 565

10
slope = 0

Magnitude (dB)
0
slope = −20 dB/dec
−10

−20
ω = 1/ T
−30
0
Phase (deg)

−45

−90
10 −2 10 −1 100 101
ω (rad/s)

Figure 7.3.5 Bode plot of the first-order system 1=ðTs þ 1Þ

(i) as ω approaches zero, M → 0, with slope = 0

(ii) as ω approaches infinity, M →  20log10 ðωT Þ, with slope = −20 dB/decade

It can be shown that these two asymptotes intersect at ω ¼ 1=T, which is called the break
frequency or corner frequency.
Because the first-order system is stable with the pole in the open left half-plane, its steady-state
response to a sinusoidal input rðtÞ ¼ r0 sin ωt exists. By Eq. (7.3.6) the steady-state response is
r0
yss ðtÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sinðωt  ψÞ (7.3.18)
1 þ ω2 T 2
with ψ ¼ tan1 ðωTÞ.
For a first-order system that is not in the standard form (7.1.18), its frequency response can
be similarly treated, as shown in Example 7.3.2.

Example 7.3.2
A phase-lag circuit is shown in Figure 7.3.6. It can be shown that the transfer function of the
circuit, from the input voltage vin to the output voltage vo , is

Vo ðsÞ R2 Cs þ 1
¼
Vin ðsÞ ðR1 þ R2 ÞCs þ 1

Let the system parameters have the following values: R1 ¼ 120 Ω; R2 ¼ 180 Ω; C ¼ 0:014 F: The
frequency response of the system then becomes


Vo ðsÞ

1 þ 2:52jω
GðjωÞ ¼
¼ :
Vin ðsÞ
1 þ 4:2jω
s¼jω
566 7 System Response Analysis

R1

R2
vin vo

Figure 7.3.6 A phase-lag circuit in Example 7.3.2

Bode Diagram
0

−1
Magnitude (dB)

−2

−3

−4

−5
0
Phase (deg)

−5

−10

−15
10−2 10−1 100 101
Frequency (rad/s)

Figure 7.3.7 Bode diagram of the phase-lag circuit in Example 7.3.2

The Bode diagram of the system frequency response is shown in Figure 7.3.7.
Furthermore, for the circuit subject to a sinusoidal input vin ðtÞ ¼ 10 cosð3tÞ, the magnitude and
phase of the system frequency response are found from the Bode diagram as follows:

A ¼ Gðj3Þ ¼ 0:6033; ϕ ¼ ∠Gðj3Þ ¼ 0:0523 rad

It follows from Eq. (7.3.8) that the steady-state output voltage of the circuit is given by

vo; ss ðtÞ ¼ 6:033 cosð3t  0:0523Þ

 7.3 System Response Analysis in the Frequency Domain 567

7.3.4 Frequency Response of Second-Order Systems

For a second-order system of the standard form (7.1.33), its frequency response is

ω2n
GðjωÞ ¼ (7.3.19)
ω2n  ω2þ 2ξωn jω

By introducing the frequency ratio

ω
r¼ (7.3.20)
ωn

the magnitude and phase of the frequency response are given by

1
A ¼ jGðjωÞj ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ð1  r2 Þ2 þ 4ξ 2 r2
 (7.3.21)
1 2ξr
ϕ ¼ ∠GðjωÞ ¼ tan
1  r2

With the magnitude and phase, the steady-state response of the system subject to a sinusoidal
input rðtÞ ¼ r0 sin ωt is determined by
r0
yss ðtÞ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sinðωt þ ϕÞ (7.3.22)
ð1  r2 Þ2 þ 4ξ 2 r2

Resonance
In Figure 7.3.8, the magnitude A and phase ϕ given in Eq. (7.3.21) are plotted against the
frequency ratio ω=ωn for different values of damping ratio ξ. The damping ratio has a
significant effect on the peak of the magnitude curve: the larger the damping ratio, the
lower the peak. The peak magnitude can be determined by dA=dr ¼ 0, which by Eq. (7.3.21)
gives r ¼ 1  2ξ2 . It follows that at the frequency
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ωr ¼ ωn 1  2ξ2 (7.3.23)

the magnitude A peaks with the value

1
Ar ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffi (7.3.24)
2ξ 1  ξ2

The parameter ωr is known as the resonant frequency because at this frequency the steady-
state amplitude of the system is significantly large. Note that Eq. (7.3.23) only makes sense
pffiffiffi
when the damping ratio falls in the range of 0 ≤ ξ ≤ 2=2. This indicates that the magnitude
568 7 System Response Analysis

Figure 7.3.8 The magnitude and phase of Eq. (7.3.21), plotted with MATLAB

pffiffiffi
A has no peak if the damping ratio is larger than 2=2, as shown in Figure 7.3.8 (for ξ = 1).
Also, by Eq. (7.3.24), the resonant magnitude Ar approaches infinity as the damping ratio ξ
vanishes. This means that small damping can cause a large amplitude, which may be
undesirable for some systems, and desirable for others, depending on the application.

Bode Diagrams
The magnitude of the Bode diagram of the system is
 7.3 System Response Analysis in the Frequency Domain 569

20

Magnitude (dB)
0
slope = 0 slope = −40 dB/dec

−20

−40

ωn
−60
0

−45
Phase (deg)

−90

−135

−180
10−1 100 101 102
ω (rad/s)
 
Figure 7.3.9 Bode plot of the second-order system Gð jωÞ ¼ ω2n = s2 þ 2ξωn s þ ω2n , plotted with
MATLAB

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
M ¼ 20log10 A ¼ 20log10 ð1  r2 Þ2 þ 4ξ 2 r2 ; r ¼ ω=ωn (7.3.25)

and the phase angle ϕ is the same as that given in Eq. (7.3.21). As shown in Figure 7.3.9, the
magnitude curve has two asymptotes:

(i) as ω approaches zero, M → 0, with slope = 0

(ii) as ω approaches infinity, M →  40log10 ðω=ωn Þ, with slope = −40 dB/decade

Also, it can be shown that the two asymptotes intersect at ω ¼ ωn , which is called break
frequency or corner frequency.

Example 7.3.3
In Figure 7.3.10, a spring–mass–damper system is subject to a base (foundation) excitation yðtÞ.
The equation of motion of the system is

_ þ kxðtÞ ¼ cyðtÞ
x ðtÞ þ cxðtÞ
m€ _ þ kyðtÞ (a)

The transfer function from the base excitation to the displacement of the mass then is

X ðsÞ cs þ k
GðsÞ ¼ ¼ 2 (b)
YðsÞ ms þ cs þ k
570 7 System Response Analysis

For a harmonic base excitation yðtÞ ¼ y0 sin ωt, the steady-state response of the system is

xss ðtÞ ¼ X sinðωt þ ϕÞ (c)

where the magnitude and phase are given by

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
k2 þ ðcωÞ2
X ¼ y0 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðk  ω2 mÞ2 þ c2 ω2 (d)
cω  cω 
ϕ ¼ tan1  tan1
k k  mω2

In the analysis and design of systems subject to base excitation, in such vehicle suspension
systems the following two transmissibility parameters are often considered.

Displacement Transmissibility The displacement transmissibility Td is a ratio of the magni-

tude of the displacement of the mass to that of the base excitation. In other words,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X k2 þ ðcωÞ2 1 þ 4ξ 2 r2
Td ¼ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (e)
y0
ðk  ω2 mÞ2 þ c2 ω2 ð1  r2 Þ2 þ 4ξ 2 r2

with the frequency ratio r ¼ ω=ωn .

Force Transmissibility The total force transmitted to the vibrating mass that is caused by
the base excitation is
   
_  yðtÞ
Ftrans ðtÞ ¼ k xðtÞ  yðtÞ  c xðtÞ _ ¼ mω2 X sinðωt þ ϕÞ (f)

Figure 7.3.10 A system subject to a harmonic base excitation

 7.4 Time Response of Linear Time-Varying and Nonlinear Systems 571

The force transmissibility Tf is a ratio of the magnitude of the transmitted force to ky0
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
FT mω2 X r2 1 þ 4ξ 2 r2 ω
Tf ¼ ¼ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; r ¼ (g)
ky0 ky0 2 ω
ð1  r2 Þ þ 4ξ r2 2 n

Figure 7.3.11 shows the displacement and force transmissibility parameters versus the frequency
ratio r for damping ratio ξ = 0.05, 0.1. 0.2, 0.5, 0.7, and 1.0, which are plotted by using Eqs. (e)
and (g). In applications, the values of the system parameters (m, c, k) are selected such that these
transmissibility parameters are in acceptable ranges.

Displacement Transmission Force Transmission

10 10
ξ = 0.05 ξ = 0.05
9 ξ = 0.1 9 ξ = 0.1
ξ = 0.2 ξ = 0.2
8 8
ξ = 0.5 ξ = 0.5
7 ξ = 0.7 7 ξ = 0.7
ξ=1 ξ=1
6 6
Td

Tf

5 5
4 4
3 3
2 2
1 1
0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
r r
(a) (b)

Figure 7.3.11 Transmissibility parameters due to base excitation: (a) displacement transmissibility;
(b) force transmissibility

7.4 Time Response of Linear Time-Varying and Nonlinear Systems

So far in this chapter, we have only considered linear time-invariant systems. In engineering
applications, however, many systems are time-varying and/or nonlinear. For example,
a spring–mass–damper system with a spring of a time-varying coefficient can be described by
 
x þ cx_ þ k þ pðtÞ x ¼ f
m€ (7.4.1)
 
where the spring force is given by fs ¼ k þ pðtÞ x, with pðtÞ being a known function of time.
As another example, a spring–mass–damper system with a nonlinear spring can be
modeled by
572 7 System Response Analysis

x þ cx_ þ k0 x þ k1 x3 ¼ f
m€ (7.4.2)

where the nonlinear spring force is given by fs ¼ k0 x þ k1 x3 , with k0 and k1 being constants.
Additionally, the rocket equation in Example 3.2.6 is another time-varying system and the
rotor system in Example 3.3.1 is another nonlinear system. For these systems, the solution
methods for system responses that are presented in the previous sections are invalid.
Analytical solutions for general linear time-varying systems and nonlinear systems are
difficult to obtain. Numerical methods are usually applied to determine the time response for
these systems. One approach to time response involves two steps: first, formulating equiva-
lent state equations, and then solving the state equations by numerical integration. Another
approach is to use software packages, such as MATLAB and Simulink. In this section, both
the approaches are demonstrated. The reader may review Appendix B and Sections 6.5 and
6.6 for the utility of MATLAB and Simulink.

7.4.1 Solution of State Equations via Numerical Integration

The governing differential equations of a dynamic system can be converted to an equivalent
matrix state equation of the form

z_ ðtÞ ¼ Φðt; zðtÞÞ; zð0Þ ¼ z0 (7.4.3)

where the state vector zðtÞ ¼ ðz1 z2 . . . zn ÞT , with n being the number of state variables.
Refer to Sections 3.7, 4.5, and 6.2 on how to derive state equations for dynamic systems. For
the linear time-varying system governed by Eq. (7.4.1), an equivalent state equation in vector
form is
!
z2 ðtÞ
z_ ðtÞ ¼ 1 (7.4.4)
ff ðtÞ  cz2 ðtÞ  ðk þ pðtÞÞz1 ðtÞg
m

For the nonlinear system described by Eq. (7.4.2), an equivalent state equation is
!
z2 ðtÞ
z_ ðtÞ ¼ 1  (7.4.5)
f ðtÞ  cz2 ðtÞ  k0 z1 ðtÞ  k1 z31 ðtÞ
m

In Eqs. (7.4.4) and (7.4.5), the state vector is zðtÞ ¼ ðz1 ðtÞ z2 ðtÞÞT ¼ ðxðtÞ xðtÞ
_ ÞT .
The state equation (7.4.3) can be solved via numerical integration algorithms. A commonly
used algorithm is the fixed-step Runge–Kutta method of order four, which is described as follows:

h
zkþ1 ¼ zk þ ð f1 þ 2f2 þ 2f3 þ f4 Þ (7.4.6)
6

See Section 2.8 for details of the numerical integration method.

 7.4 Time Response of Linear Time-Varying and Nonlinear Systems 573

7.4.2 Solution by MATLAB and Simulink

Equation (7.4.3) can be solved using the MATLAB function ode45, which is a Runge–
Kutta solver with an automatic step-size adjustment capability. Also, the software Simulink
can be used to build up a model according to Eq. (7.4.3), for simulation of system response.

Example 7.4.1
Consider the time-varying system described by Eq. (7.4.1), with the following parameters

m ¼ 1; c ¼ 1:2; k ¼ 16; pðtÞ ¼ 0:3 sinð3:8tÞ

Here, for convenience, nondimensional values of the system parameters have been assigned.
Determine the time response of the system subject to a step input f ðtÞ ¼ 16 and zero initial
displacement and velocity. Compare the result with the step response of the corresponding time-
invariant system, which is governed by Eq. (7.4.1) with pðtÞ ¼ 0.
Solution
From Eq. (7.4.4), the equivalent state equation is
 
z2 0
z_ ¼ ; with zð0Þ ¼
16  1:2z2  ð16 þ 0:3 sinð3:9tÞÞz1 0

The displacement xðtÞ of the time-varying system, which is the first element of the state vector zðtÞ,
is computed by the fixed-step Runge–Kutta method and plotted in Figure 7.4.1. For comparison,
the step response yðtÞ of the corresponding time-invariant system

1.8 1.8
x(t) x(t)
1.6 y(t) 1.6 y(t)

1.4 1.4

1.2 12
Displacement

Displacement

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 1 2 3 4 5 6 7 8 9 10 0 2 4 6 8 10 12 14 16 18 20
Time, t Time, t
(a) (b)

Figure 7.4.1 Figure 7.4.1 Step responses: solid line – the time-varying system; dashed line – the
time-invariant system for (a) 0 ≤ t ≤ 10 and (b) 0 ≤ t ≤ 20
574 7 System Response Analysis


σ
yðtÞ ¼ 1  eσt cos ωd t þ sin ωd t
ωd

with σ ¼ 0:6 and ωd ¼ 3:9547 is also plotted in the figure. Here, yðtÞ is obtained by using Eq.
(7.1.57). As seen from Figure 7.4.1(a), the displacement of the time-invariant system settles at its
steady-state value ðyss ¼ 1Þ before t ¼ 10. However, the displacement of the time-varying system,
due to the sinusoidal variation pðtÞ of the spring coefficient, remains of harmonic type, as shown in
Figure 7.4.1(b).
The MATLAB scripts for plotting Figures 7.4.1(a) and (b) are as follows:
% Example 7.4.1
% Time-varying system
eta0 = [0; 0];
tf = 10; npts = 4001; h = tf/(npts-1);
t = linspace(0, tf, npts);
z = zeros(2, npts);
z(:,1) = eta0;
y = zeros(1, npts);
for i = 1: npts-1
tt = t(i); zz = z(:,i);
f1 = force_exm741(tt, zz);
f2 = force_exm741(tt+h/2, zz+h/2*f1);
f3 = force_exm741(tt+h/2, zz+h/2*f2);
f4 = force_exm741(tt+h, zz+h*f3);
z(:,i+1) = zz +h/6*(f1+2*f2+2*f3+f4);
end
x = z(1,:);
% Time-invariant system
sgm = 0.6; wd = sqrt(16*4-1.2^2)/2;
y = 1-exp(-sgm*t).*(cos(wd*t)+sgm/wd*sin(wd*t));
% Plotting
plot(t,x,t,y,’– –’)
grid
xlabel(‘Time, t’)
ylabel(‘Displacement’)
legend(‘x(t)‘, ‘y(t)’)
% A function for the time-varying system
function f = force_exm741(tt,z)
m = 1; c = 1.2; k = 16;
 7.4 Time Response of Linear Time-Varying and Nonlinear Systems 575

e0 = 0.3; omg = 3.8;

f0 = 16;
f1 = z(2);
f2 = (f0-c*z(2)-(k+e0*sin(omg*tt))*z(1))/m;
f = [f1; f2];

Example 7.4.2
Consider the nonlinear system described by Eq. (7.4.2). Let the system parameters have the non-
dimensional values: m ¼ 1; c ¼ 1:2; k0 ¼ 16; k1 ¼ 0:5. Assume zero initial disturbances. For
a constant external force (step input) f ðtÞ ¼ 16, determine the forced response of the system by
Simulink.

Solution
A Simulink model of the nonlinear system is established in Figure 7.4.2. Refer to Section 6.6 for
how to build a Simulink model. Running this model yields the forced response xðtÞ plotted in
Figure 7.4.3. For comparison, the response yðtÞ of the corresponding linear system ðk1 ¼ 0Þ is also
plotted in the figure. As observed from the figure, the linear system has a large amplitude in
response. This is because the term k1 x3 in the spring force of the nonlinear system has a hardening
effect.

Figure 7.4.2 Simulink model of the nonlinear mechanical system in Example 7.4.2
576 7 System Response Analysis

1.8
x(t)
1.6 y(t)
1.4

1.2
Displacement
1

0.8

0.6

0.4

0.2

0
0 1 2 3 4 5 6 7 8 9 10
Time, t

Figure 7.4.3 Step responses: solid line – nonlinear system; dashed line – linear system

MATLAB Function Blocks for Nonlinear State Equations

In Simulink, the Used-Defined Functions library has a MATLAB Function block, with
the icon , which can be used to determine the solutions of time-varying and/or
nonlinear state equations.

Example 7.4.3
Repeat the same problem in Example 7.4.2 by using the MATLAB Function block.

Solution
With the assigned parameter values and the constant input f ðtÞ ¼ 16, the nonlinear state equations
(7.4.5) become
 
z_ 1 ðtÞ z2 ðtÞ
¼
z_ 2 ðtÞ 16  1:2z2 ðtÞ  16z1 ðtÞ  0:5z31 ðtÞ

Through use of the MATLAB Function block, a Simulink model is created in Figure 7.4.4.
Double-click the MATLAB Function block shows the MATLAB commands as follows:
function [y1, y2] = fcn(z1, z2)
y1 = z2;
y2 = 16-1.2*z2-16*z1-0.5*z1^3;
Running simulation with the model gives the same x(t) curve as shown in Figure 7.4.3.
 Chapter Summary 577

Figure 7.4.4 Simulink model using the MATLAB Function block

CHAPTER SUMMARY
This chapter addresses three important issues regarding modeling, analysis, and design of
dynamic systems: system response analysis in the time domain, stability analysis, and system
response analysis in the frequency domain. In time-domain analysis, system order (e.g., first
order, second order, and higher orders), input type (e.g., impulse, step, ramp, and sinusoidal
functions), free response and forced response, and transient response and steady-state response
are classified. Analytical solutions of the impulse and step responses for first- and second-order
systems are presented. In particular, four types of damping cases (no damping, underdamping,
critical damping, and overdamping) of second-order systems are thoroughly examined, and
the effects of damping on time response are illustrated. Also, the step response of an
underdamped second-order system is evaluated and quantified through use of performance
specification parameters. MATLAB and Simulink are used to compute time responses for
higher-order systems, time-varying systems, and nonlinear systems.
In stability analysis, the concept of stability is first introduced; three stability definitions
based on free response or impulse response are then provided; and afterwards stability
conditions for a linear time-invariant system are related to the location of its transfer
function poles. To detect the stability of a linear time-invariant system without having to
know its transfer function poles, the Routh–Hurwitz stability criterion is introduced.
In system response analysis in the frequency domain, the steady-state behavior of
a dynamic system subject to a sinusoidal input is described and the frequency response
function of the system is defined. For a given dynamic system, the frequency response plots
of M and ϕ against ω are derived; Bode diagrams and Nyquist plots are introduced; and the
use of MATLAB in plotting frequency-response plots is demonstrated.
578 7 System Response Analysis

Upon completion of this chapter, you should be able to:

(1) Obtain the impulse and step responses of first- and second-order systems in analytical
form.
(2) Determine the parameter values of an underdamped second-order system that meet the
performance specifications on its step response in terms of rise time, maximum over-
shoot, and settling time.
(3) Compute the time response of general dynamic systems, including linear time-invariant,
time-varying, and nonlinear systems, by numerical integration and via MATLAB and
Simulink.
(4) Examine the stability of linear time-invariant systems by two methods: (i) examining
transfer function pole locations; and (ii) applying the Routh–Hurwitz stability criterion.
(5) Determine the steady-state response of linear time-invariant systems.
(6) Plot the frequency response of linear time-invariant systems, by either analytical formulas
or MATLAB.

REFERENCES
1. R. C. Dorf and R. H. Bishop, Modern Control Systems, 13th ed., Pearson, 2016.
2. G. Franklin, J. Powell, and A. Emami-Naeini, Feedback Control of Dynamic Systems, 8th ed.,
Pearson, 2018.
3. N. S. Nise, Control Systems Engineering, 8th ed., Wiley, 2020.

PROBLEMS
Section 7.1 System Response Analysis in the Time Domain
7.1 A dynamic system is governed by the differential equation
 
D2 þ 2D þ 4 y ¼ 3ð1  e5t Þ; D ¼ d=dt

and the initial conditions

_
yð0Þ ¼ 0:1; yð0Þ ¼ 1:

(a) Plot the free response of the system.

(b) Plot the forced response of the system.
Hint: You may use MATLAB or Simulink to plot the system response.
7.2 A dynamic system is governed by the differential equation
 
D3 þ 4D2 þ 2D þ 3 y ¼ 12; D ¼ d=dt
 Problems 579

and the initial conditions

_
yð0Þ ¼ 0:5; yð0Þ ¼ 0; y€ð0Þ ¼ 2:

(a) Plot the free response of the system.

(b) Plot the forced response of the system.
Hint: You may use MATLAB or Simulink to plot the system response.
7.3 For each of the following models, check if the condition of the final-value theorem is
satisfied. If the condition is met, evaluate the steady-state response of the system by the
final-value theorem. If the condition is not satisfied, plot the system response by
MATLAB, to show that no steady-state value is reached.
 
(a) D2 þ 2D þ 4 y ¼ 3ð1  e5t Þ; D ¼ d=dt
yð0Þ ¼ 0; yð0Þ _ ¼1
 
(b) D2  2D þ 4 y ¼ 3ð1  e5t Þ; D ¼ d=dt
yð0Þ ¼ 1; yð0Þ _ ¼0
 3 
(c) D þ 4D þ 2D þ 3 y ¼ 10; D ¼ d=dt
2

yð0Þ ¼ 1 yð0Þ _ ¼ 0; y€ð0Þ ¼ 0

 
(d) D3 þ 2D2 þ 3D þ 6 y ¼ 10; D ¼ d=dt
_
yð0Þ ¼ 1; yð0Þ ¼ 0; y€ð0Þ ¼ 0
7.4 For an armature-controlled DC motor in Section 6.4, its governing equations are

dω
J þ cb ω ¼ Kia  TL (i)
dt
dia
La þ Ra ia ¼ vin  Kb ω (ii)
dt

where ω is the rotation speed of the motor; vin and ia are the voltage input and current of
the armature circuit, respectively; K is the motor torque constant; Kb is the back
electromotive force constant; and TL denotes a load torque. Consider step inputs:
vin ðtÞ ¼ V0 and TL ¼ τ0 , where V0 is a constant voltage and τ0 is a constant torque.
Assume that KV0 > Ra τ0 . Let the governing equations have zero initial conditions.
Determine the steady-state rotation speed of the motor, ωss ¼ lim ωðtÞ.
t→∞
7.5 Figure P7.5 shows a dynamic system in a block diagram, which has two inputs and one
output. Consider the step inputs: rðtÞ ¼ 5 and dðtÞ ¼ 3. Assume zero initial conditions.
(a) Determine the steady-state output of the system, yss ¼ lim yðtÞ. Hint: You may use
t→∞
the final-value theorem.
(b) Plot the system output y(t) by MATLAB. Show that the steady-state response in the
simulation is the same as predicted in part (a).
580 7 System Response Analysis

D(s)

R(s) + + Y(s)
+ 1
3s + 4
− s2 (s + 8)

Figure P7.5

7.6 Consider the following three first-order systems:

System I: x_ þ 3x ¼ r
System II: 4x_ þ 3x ¼ r
System III: 4x_ þ 3x ¼ 2_r þ r
Every system is subject to an impulse input rðtÞ ¼ δðtÞ, where δðtÞ is the Dirac delta
function. Assume zero initial conditions.
(a) Obtain the impulse responses of the system by Laplace transform.
(b) Plot the impulse responses of the systems versus time by MATLAB.
7.7 Consider the following first-order systems
System I: x_ þ x ¼ r
System II: 2x_ þ x ¼ r
System III: 5x_ þ x ¼ r
Every system is subject to a unit step input. Assume zero initial conditions.
(a) Obtain the step response of each system by Laplace transform.
(b) Plot the step responses of the three systems versus time in one figure by MATLAB
and discuss the effect of the time constant on the system response.
7.8 An RC circuit of resistance R and capacitance C is governed by
ð
1
Ri þ idt ¼ vin
C
ð
1
vout ¼ idt
C

where vin is the input voltage and vout is the output voltage (capacitor voltage or system
response). Let the input voltage be a constant, vin ðtÞ ¼ E0 . Assume that the initial output
voltage is vout ð0Þ ¼ Vc0 . Determine the free, forced, and the steady-state responses of the
circuit.
7.9 According to Section 5.6, a single-tank liquid-level system, with constant cross-section
area A, liquid height h, volume inflow rate qin , and fluid resistance R, is governed by

dh g
A þ h ¼ qin
dt R
 Problems 581

where g is the gravitational acceleration. Let the initial liquid height be hð0Þ ¼ h0 .
Assume a step inflow rate, qin ¼ q0 . Determine the total response (h) of the liquid-level
system. Also find the steady-state liquid head hss .
7.10 Consider a first-order system in the standard form of Eq. (7.1.17). Under a step input
rðtÞ ¼ r0 and with the initial condition yð0Þ ¼ 2r0 , the system response at t = 2 s is
yð2Þ ¼ 1:5r0 .
(a) Determine the time constant T of the system.
(b) With the time constant of part (a), determine the time for the system output y to
reach a 5% neighborhood of its steady-state value yss .
7.11 For the following second-order systems, consider the following initial conditions:
_
xð0Þ ¼ 1; xð0Þ ¼ 0: Determine the natural frequency and damping ratio for each
system. Also, plot the free responses of all the systems versus time by MATLAB.
(a) x€ þ 4x ¼ 0
(b) x€ þ x_ þ 4x ¼ 0
(c) x€ þ 4x_ þ 4x ¼ 0
(d) x€ þ 5x_ þ 4x ¼ 0
7.12 For the following second-order systems, consider an impulse input rðtÞ ¼ δðtÞ, with δðtÞ
being the Dirac delta function. Determine the natural frequency and damping ratio for
each system. Also, plot the impulse responses of all the systems versus time by
MATLAB. Zero initial conditions are assumed for each system.
(a) x€ þ 4x ¼ r
(b) x€ þ x_ þ 4x ¼ r
(c) x€ þ 4x_ þ 4x ¼ r
(d) x€ þ 5x_ þ 4x ¼ r
7.13 For the following second-order systems, consider a step input rðtÞ ¼ 4uðtÞ, with uðtÞ
being the unit step function. Determine the damping characteristics (undamped,
underdamped, critically damped, or overdamped) of each system. Also, plot the step
responses of the systems versus time by MATLAB. Zero initial conditions are assumed
for each system.
(a) x€ þ 4x ¼ r
(b) x€ þ x_ þ 4x ¼ r
(c) x€ þ 4x_ þ 4x ¼ r
(d) x€ þ 5x_ þ 4x ¼ r
7.14 Consider the L-shaped rigid bar in Figure P7.14. Assume small rotation, jθj ≪ 1.
(a) Determine the natural frequency and damping ratio of the bar.
(b) Derive the conditions on the spring and damping coefficients (k, c) and the
geometric parameters of the bar for the bar to be overdamped.
582 7 System Response Analysis

k
A

C f (t )

h
a

θ
B
O

c
l

Figure P7.14

7.15 A series RLC circuit is shown in Figure P7.15, where the applied voltage vin is the input
and the capacitor voltage vC is the output.
(a) Show that the circuit is a second-order system.
(b) Determine the natural frequency and damping ratio of the circuit in terms of
parameters R, L, and C.
(c) For a constant voltage, vin ¼ E0 , find the steady-state output of the circuit.

R L

+
vin C vc
i
−

Figure P7.15

7.16 Consider the feedback control system in Figure P7.16, in which ω > 0.
(a) Determine the conditions of parameters a and b such that the system is
underdamped.
(b) With the condition derived in part (a), determine if the system has a steady-state
response to a step input, rðtÞ ¼ r0 . If so, find yss ¼ lim yðtÞ.
t→∞

R(s) + 1 Y(s)
as + b
− s2 + ω2

Figure P7.16
 Problems 583

7.17 For the following second-order systems, consider a step input rðtÞ ¼ 25uðtÞ, with uðtÞ
being the unit step function. Compute the 100% rise time, peak time, maximum
overshoot, and settling time (with a 2% criterion) for each system. Also, plot the step
responses of the systems versus time in one figure.
(a) x€ þ 2x_ þ 25x ¼ r
(b) x€ þ 5x_ þ 25x ¼ r
(c) x€ þ 8x_ þ 25x ¼ r
7.18 Consider a dynamic system, with its response y governed by the differential equation

x þ ð3 þ aÞx_ þ ð18 þ bÞx ¼ r

2€

where r is an input.
(a) Determine the values of a and b such that the natural frequency and damping ratio
of the system are ωn ¼ 4 and ξ ¼ 0:5, respectively.
(b) With a and b determined in part (a), compute the 100% rise time, maximum
overshoot, and settling time (with a 5% criterion) of the step response of the system.
(c) With the values of a and b from part (a), plot the step response of the system by
MATLAB, for rðtÞ ¼ 32.
7.19 Consider the L-shaped rigid bar in Figure P7.14, which is in small rotation, jθj≪1.
Assume that h = 3 m, l = 4 m, a = 2 m, and IO = 100 kg·m2.
(a) Determine the values of the spring and damping coefficients (k, c), such that the
roots of the characteristic equation of the system are 2  j9.
(b) With the k and c from part (a), compute the 100% rise time, maximum overshoot,
and settling time (with a 2% criterion) of the step response of the system.
(c) Use MATLAB to plot the step response of the system, with f ðtÞ ¼ 12 N.
7.20 Figure P7.20 shows a dynamic system in closed-loop format.
(a) Determine parameters a and b such that the step response of the system meets the
following two performance specifications: (i) Mp ≤ 15%; and (ii) ts = 1 s with a 1%
criterion.
(b) Taking the values of a and b from part (a), plot the system response subject to a step
input rðtÞ ¼ 1.

R (s) + 1 Y (s)

− s(s + 4)

as + b

Figure P7.20
584 7 System Response Analysis

7.21 A second-order system is described by the transfer function

YðsÞ 20ðs=a þ 1Þ
GðsÞ ¼ ¼
RðsÞ s2 þ 4s þ 20

where a > 0. The system can be viewed as a standard second-order system with an
added zero: z ¼ a. When a → ∞, the transfer function becomes that for the
standard second-order system described by Eq. (7.1.31). Let the input be a unit step
function. In one figure, plot the step responses of the system for a = 2, 5, 10, 20, and ∞
and for 0 ≤ t ≤ 3 s: Discuss the effect of the added zero on the system response.
7.22 A third-order system is described by the transfer function
YðsÞ 25
GðsÞ ¼ ¼ 2
RðsÞ ðs þ 6s þ 25Þðs=b þ 1Þ

where b > 0. The system can be viewed as a standard second-order system with an
added pole: p ¼ b. When b → ∞, the transfer function becomes that for the
standard second-order system described by Eq. (7.1.31). Let the input be a unit step
function. In one figure, plot the step responses of the system for b = 3, 6, 15, 30, and ∞
and for 0 ≤ t ≤ 2:5 s: Discuss the effect of the added pole on the system response.
X ðsÞ 4ð0:5s þ 1Þ
7.23 For a system ¼ 2 subject to a step input rðtÞ ¼ 1, obtain
FðsÞ ðs þ 1:2s þ 4Þð0:2s þ 1Þ
a mathematical expression of the forced response and plot the response. Hint: follow
Example 7.1.10 and apply partial fraction expansion.

Section 7.2 Stability Analysis

7.24 Listed below are the characteristic equations of six dynamic systems.
System 1: s2 þ 9 ¼ 0
System 2: s2 þ 2s þ 9 ¼ 0
System 3: s2  2s þ 9 ¼ 0
System 4: s3 þ 2s2 þ 5s þ 9 ¼ 0
System 5: s3 þ 4s2 þ 9s þ 36 ¼ 0
System 6: s4 þ 18s2 þ 81 ¼ 0
Obtain the poles for each system by MATLAB and determine the stability of the
system.
7.25 Listed below are the characteristic equations of three dynamic systems.
System 1: s3 þ 3s2 þ 5s þ 7 ¼ 0
System 2: s3 þ 6s2  36s þ 40 ¼ 0
System 3: s4 þ 3s3 þ 5s2 þ 7s þ 9 ¼ 0
 Problems 585

By the Routh–Hurwitz stability criterion, examine the stability of each system. Also,
for an unstable system, determine the number of unstable poles (poles with positive real
parts) without solving the characteristic equation.
7.26 Consider the characteristic equations of the following two systems:
System 1: s3 þ 6s2 þ ða  36Þs þ ð40 þ bÞ ¼ 0
System 2: s4 þ s3 þ 3s2 þ as þ b ¼ 0
For each system, determine the conditions on a and b such that the system is stable.
7.27 A feedback control system is shown in Figure P7.27, where a third-order plant (a
system under control) is controlled by a PID controller (Section 8.3).
(a) Determine whether or not the plant is stable.
(b) Determine the conditions on the gain constants (KP , KI , KD ) such that the feedback
control system is stable.

PID controller Plant

R (s) + 1 1 Y(s)
Kp + KI + K Ds
− s s 2 (s + 4)

Figure P7.27

7.28 In Figure P7.28, a field-controlled DC motor is under feedback control, where ΩðsÞ is
the motor speed; RðsÞ is the reference input; K, J, cb , Lf , and Rf are the motor
parameters defined in Section 6.4; KP and KI are the gain constants of a PI controller
(Section 8.3); ga is the gain constant of the power amplifier; and gs is the gain constant
of the motor speed sensor. Determine the conditions of the controller gain constants
for the control system to be stable.

Controller DC motor
R(s ) + E(s ) 1 Vin (s ) K Ω (s )
Kp + KI ga
− s (Js + cb)(Lf s + Rf )
Amplifier

gs
Sensor

Figure P7.28
586 7 System Response Analysis

7.29 For the thermal control system in Figure 8.7.3, its characteristic equation is
 
sðRCs þ 1Þðτa s þ 1Þ þ Rga gs KP s þ KI þ KD s2 ¼ 0

where C and R are the thermal capacitance and resistance of the container (plant),
respectively; gs is the gain constant of the sensor; τa and ga are the time constant and
gain constant of the heater (actuator); and KP , KI , and KD are the gain constants of the
PID controller (Section 8.3). Determine the conditions of the controller gain constants
such that the thermal control system is stable.

Section 7.3 System Response Analysis in the Frequency Domain

7.30 Determine the steady-state response yss ðtÞ of each of the following first-order systems to
a sinusoidal input.
YðsÞ 2
System 1: ¼ ; rðtÞ ¼ 10 sinð2tÞ
RðsÞ s þ 16
YðsÞ s þ 16
System 2: ¼ ; rðtÞ ¼ 2 cosð10t þ 30°Þ
RðsÞ s þ 25
YðsÞ 2s  3
System 3: ¼ ; rðtÞ ¼ 4 sinð5tÞ þ 3 cosð5tÞ
RðsÞ sþ8
7.31 Determine the steady-state response yss ðtÞ of each of the following second-order
systems to a sinusoidal input.
YðsÞ 2
System 1: ¼ ; rðtÞ ¼ 10 sinð3tÞ
RðsÞ s2 þ 2s þ 10
YðsÞ 2ðs þ 1Þ
System 2: ¼ ; rðtÞ ¼ 5 cosð4t þ 45°Þ
RðsÞ s2 þ 3s þ 16
YðsÞ 4s  1
System 3: ¼ ; rðtÞ ¼ 4 sinð6tÞ þ 3 cosð6tÞ
RðsÞ s2 þ 5s þ 25
7.32 Consider the L-shaped rigid bar Figure P7.14, which is in small rotation, jθj ≪ 1.
Assume the parameters of the bar as follows: h = 3 m, l = 4 m, a = 2 m, IO =
100 kg·m2, k = 50 N/s, and c = 8 N·s/m. Let the external force be f ðtÞ ¼ 5 sinðωtÞN,
where ω is the excitation frequency.
(a) Plot the magnitude and phase of the steady-state response θss ðtÞof the bar, versus
the excitation frequency ω.
(b) Following Section 7.3.4, determine the resonant frequency ωr of the bar.
7.33 Consider the series RLC circuit in Figure P7.15, with R = 0.1 Ω, L = 0.01 H, and C = 1
F. Let the input voltage be vin ¼ 12 cosðωtÞ V.
(a) Plot the magnitude and phase of the steady-state response vC;ss ðtÞ of the circuit.
(b) Following Section 7.3.4, determine the resonant frequency ωr of the circuit.
 Problems 587

7.34 Consider the control system in Figure P7.20, with a > 0 and b > 0.
(a) For a sinusoidal input rðtÞ ¼ r0 sinðωtÞ, obtain the expression of the steady-state
output yss ðtÞ of the system.
(b) For a = 1 and b = 400, plot the magnitude and phase of yss ðtÞ.
7.35 Use MATLAB to plot the Bode diagrams of the following systems:
YðsÞ 5ðs þ 1Þ
System 1: ¼
RðsÞ s þ 10
YðsÞ 2s
System 2: ¼ 2
RðsÞ s þ 3s þ 16
YðsÞ 4ðs þ 5Þ
System 3: ¼
RðsÞ s3 þ 4s2 þ 9s þ 36
7.36 Use MATLAB to plot the Nyquist plots of the following systems:
YðsÞ 5ðs þ 1Þ
System 1: ¼
RðsÞ s þ 10
YðsÞ 2s
System 2: ¼ 2
RðsÞ s þ 3s þ 16
YðsÞ 4ðs þ 5Þ
System 3: ¼ 3
RðsÞ s þ 4s2 þ 10s þ 36

Section 7.4 Time Response of Linear Time-Varying and Nonlinear Systems

7.37 Consider the cart–pendulum system in Figure P7.37, which is a nonlinear dynamic system.
The system parameters are given as follows: M = 20 kg, mb ¼ 5 kg, k = 500 N/m,
c = 20 N·s/m, and L = 1.2 m. A sinusoidal force is applied to the mass
M: f ¼ 10 sinð6tÞ N. The initial conditions of the system are as follows: yð0Þ ¼ 0;
_
yð0Þ _
¼ 0; θð0Þ ¼ 0:02 rad; θð0Þ ¼ 0.

c k

M
g
y
f
L
θ

mb

Figure P7.37
588 7 System Response Analysis

By the fixed-step Runge–Kutta method of order four, as shown in Section 7.4, compute
and plot the response (y, θ) of the system for 0 ≤ t ≤ 10 s. Hint: make sure to select
a small enough step size h.
7.38 Repeat Problem 7.37 by building a Simulink model. Hint: you may follow
Example 7.4.3.
7.39 Consider the dynamic system governed by the following differential equations

y 1 þ 3y_ 1 þ ð18 þ 2 sin 3tÞy1  3y_ 2  8y2 ¼ f

2€
y 2 þ 3y_ 2 þ 8y2 þ y32  3y_ 1  8y1 ¼ 0
5€

which is nonlinear and time-variant. Let the input of the system be

f ¼ 2ð1  e0:5t Þsin4t. Assume the initial conditions as follows: y1 ð0Þ ¼ 0:2;
y_ 1 ð0Þ ¼ 0; y2 ð0Þ ¼ 0; y_ 2 ð0Þ ¼ 1. By the fixed-step Runge–Kutta method of order
four, as shown in Section 7.4, compute and plot the dynamic response (y1 and y2 ) for
0 ≤ t ≤ 8 s.
7.40 Repeat Problem 7.39 by using the MATLAB function ode45.
8 Introduction to Feedback Control Systems

Contents
8.1 General Concepts 589
8.2 Advantages of Closed-Loop Control Systems 595
8.3 PID Control Algorithm 601
8.4 Control System Analysis 611
8.5 The Root Locus Method 619
8.6 Analysis and Design by the Root Locus Method 632
8.7 Additional Examples of Control Systems 648
Chapter Summary 653
References 654
Problems 654

Controls in engineering are efforts to change, design, or modify the behaviors of dynamic
systems. Automatic control is control that involves only machines and devices, and that has
no human intervention. Examples of automatic control are diverse, including room-tem-
perature control, cruise control of cars, missile guidance, trajectory control of robots,
control of appliances such as washing machines and refrigerators, and control of industrial
processes like papermaking and steelmaking. In this chapter, for simplicity, an automatic
control system is called a control system. Because the focus of this text is on the modeling and
analysis of dynamic systems, only basic concepts of feedback control are introduced. For
theories and methods about feedback control systems, one may refer to standard textbooks
for control courses, including those listed at the end of the chapter.

8.1 General Concepts

A control system usually consists of the following four major components or subsystems:

Plant: A plant is a dynamic system that is controlled so that its output meets certain
performance requirements. The output of the plant is also the output of the control

589
590 8 Introduction to Feedback Control Systems

system. In engineering applications, a plant can be a machine, a device, a structure, or an

industrial process.
Sensor: A sensor is a device that measures the output of the plant. Examples of sensors in
engineering applications include accelerometers, tachometers, displacement transducers,
strain gauges, manometers, flowmeters, magnetometers, and thermometers. As will be
explained, not every control system requires a sensor.
Controller: A controller is a component that computes a control signal or control command
by an algorithm. The algorithm, which is also called control logic or control law, makes
use of the specified performance requirements and the measured plant output if necessary.
Nowadays, a controller is usually implemented on a computer chip (logic board).
Actuator: An actuator is a device that applies a control effort (e.g., a force, a voltage, or a
heat flow rate) to the plant according to the control command from the controller.

Depending on whether the plant output is measured, there are two control system
configurations: open-loop control systems and closed-loop control systems, which are
explained as follows.

Open-Loop Control System

An open-loop control system (simply called an open-loop system) is a control system in
which the plant is controlled without measurement of the plant output; see Figure 8.1.1(a). In
other words, no sensor is implemented in an open-loop system and the output of the control
system has no influence on the control signal of the controller.

Closed-Loop Control System

A closed-loop control system is a control system in which the plant is controlled via
measurement of the plant output and comparison of the measured output with a reference
input (desired output); see Figure 8.1.1(b). A closed-loop control system is also called a
closed-loop system, feedback control system, or simply feedback system. In contrast to open-
loop control, the output of a closed-loop system has influence on the control signal from the
controller. In such a control system, the sensor data are fed back to close a loop from the
output to the reference input. Because of this, a closed-loop system is also called a feedback
control system.
Control systems are seen commonly in everyday life. An electric toaster in the kitchen is an
example of an open-loop control system; an air conditioner is an example of a closed-loop
control system. Open-loop control, which does not require an output measurement, is easier
and less expensive to implement. However, the performance of an open-loop control system
may not be satisfactory. Due to the complexity of dynamic systems in many applications and
because of the need for optimal performance, closed-loop control plays an essentially important
role in the development and advancement of modern technology and civilization.
 8.1 General Concepts 591

Disturbance

Reference Control Control

input signal action Output
Controller Actuator Plant

(a)

Disturbance

Reference Control Control

input + Error signal action Output
Controller Actuator Plant
−

Sensor

Sensor noise

(b)

Figure 8.1.1 Control system configurations: (a) open-loop control system; and (b) closed-loop control
system

Closed-loop control has three major advantages over open-loop control: (i) it can reject a
disturbance and sensor noise; (ii) it reduces the influence of the variations of system param-
eters; and (iii) it can improve the transient and steady-state response and stability of a control
system. These advantages are illustrated in Section 8.2.

Example 8.1.1
Figure 8.1.2 shows a closed-loop system for temperature control of a house in hot weather, by an
air conditioner. In this control system, the temperature of a room is measured by a thermostat. A
controller (logic board) compares the measured room temperature with a desired temperature
(preset as a reference input) and computes a control signal. Based on the control signal (the
controller output), a trio of compressor, condenser, and evaporator as the actuator removes a
certain amount of heat from the house via a working fluid. This process continues until the room
temperature reaches the preset temperature as desired. The thermostat, controller, compressor,
condenser, and evaporator are the parts of the air conditioner.
592 8 Introduction to Feedback Control Systems

Air conditioner
Heat
Comparison
Desired Room
Compressor temperature
temperature + Error
Controller Condenser House
− Evaporator

Thermostat

Figure 8.1.2 Block diagram of a room-temperature control system enabled by an air conditioner

Transfer Function Formulation of Control Systems

Assume that all the components of a control system can be described as linear time-invariant
subsystems. The two control system configurations in Figure 8.1.1 then can be presented in the s-
domain block diagrams shown in Figure 8.1.3, where R(s) is the reference input, Y(s) is the output,
E(s) is the error (for closed-loop control only), U(s) is the control signal or controller output, D(s)
is a disturbance applied to the plant, N(s) is an sensor noise, and Gp ðsÞ; Gs ðsÞ; Gc ðsÞ; Ga ðsÞ are
the transfer functions of the plant, sensor, controller, and actuator, respectively.
The output of the open-loop system, by Figure 8.1.3(a), is written as

YðsÞ ¼ Gp ðsÞGa ðsÞGc ðsÞRðsÞ þ Gp ðsÞDðsÞ (8.1.1)

As can be seen from the previous equation, the controller Gc ðsÞ has no influence on the
disturbance D(s). This disadvantage is also demonstrated through an example (motor speed
control) in Section 8.2.

D (s) Plant
Controller Actuator
R (s) U (s) V (s) ++ Y (s)
Gc (s) Ga (s) Gp (s)

(a)

D(s)
Controller Actuator Plant
R (s) + E (s) U (s) V (s) + Y (s)
Gc (s) Ga (s) + Gp (s)
−

Gs (s)
+
+
Sensor
N (s)
(b)

Figure 8.1.3 Block diagrams of control systems: (a) open-loop; and (b) closed-loop
 8.1 General Concepts 593

The output of the closed-loop control system, Figure 8.1.3(b), is of the form

YðsÞ ¼ TR ðsÞRðsÞ þ TD ðsÞDðsÞ þ TN ðsÞNðsÞ (8.1.2)

where TR ðsÞ, TD ðsÞ, and TN ðsÞ are the transfer functions describing the influence of the
reference input, disturbance, and sensor noise on the output of the closed-loop control
system, respectively. In other words, the transfer functions can be expressed as




YðsÞ

YðsÞ

YðsÞ


TR ðsÞ ¼
; TD ¼
; TN ðsÞ ¼
(8.1.3)
RðsÞ
DðsÞ
NðsÞ

DðsÞ ¼ 0 RðsÞ ¼ 0 RðsÞ ¼ 0
N ðsÞ ¼ 0 N ðsÞ ¼ 0 DðsÞ ¼ 0

With the formulas of the block diagrams in Section 6.2.3, these transfer functions are found
to be

Gc ðsÞGa ðsÞGp ðsÞ Gp ðsÞ LðsÞ

TR ðsÞ ¼ ; TD ðsÞ ¼ ; TN ðsÞ ¼  (8.1.4)
1 þ LðsÞ 1 þ LðsÞ 1 þ LðsÞ

with
LðsÞ ¼ Gs ðsÞGc ðsÞGa ðsÞGp ðsÞ (8.1.5)

Here, LðsÞ is known as the loop transfer function. As can be seen from Eqs. (8.1.4) and (8.1.5),
in a closed-loop control system, the controller Gc ðsÞ has influence on the disturbance DðsÞ
and noise NðsÞ. This is one main reason that closed-loop control systems are widely used
in engineering applications. Accordingly, this chapter focuses on closed-loop control
systems.

Example 8.1.2
The displacement of a mass–damper system is under feedback control; see Figure 8.1.4(a), where f
is an external force and fc is the control force applied to the mass by the actuator. Without loss of
generality, assume that the sensor measures the displacement of the mass by ys ðtÞ ¼ gs xðtÞ and that
the actuator force is related to the control signal u by fc ðtÞ ¼ ga uðtÞ, where gs and ga are gain
constants. A proportional plus derivative (PD) controller, which is presented in detail in Section
8.3, is implemented. The transfer function of the plant (the mass–damper system) without control
is given by

X ðsÞ 1
¼
FðsÞ ms2 þ cs
The PD controller is described by the transfer function

UðsÞ
¼ KP þ KD s
EðsÞ
594 8 Introduction to Feedback Control Systems

Figure 8.1.4 A mass–damper system under PD feedback control in Example 8.1.2

where KP and KD are the gain constants of the controller, and EðsÞ is the Laplace transform of the
error e ¼ r  ys , with r being a reference input.
With the above description, a block diagram with component transfer functions is established
in Figure 8.1.4(b). By the block diagram and from Eq. (8.1.2), the s-domain output of the closed-
loop system is obtained as

X ðsÞ ¼ TR ðsÞRðsÞ þ TD ðsÞFðsÞ

in which the transfer functions are given by

1
ðKP þ KD sÞga 2 þ cs ga ðKD s þ KP Þ
TR ðsÞ ¼ ms ¼ 2
1 ms þ ðc þ gs ga KD Þs þ gs ga KP
1 þ gs ðKP þ KD sÞga 2
ms þ cs
1
TD ðsÞ ¼ ms2 þ cs ¼ 2
1
1 ms þ ðc þ gs ga KD Þs þ gs ga KP
1 þ gs ðKP þ KD sÞga 2
ms þ cs

As indicated in Eq. (8.1.4), the two transfer functions have the same denominator

ms2 þ ðc þ gs ga KD Þs þ gs ga KP

In control system design, the control gains KP and KD are selected such that the system output meets
certain stability and performance specifications. This will be discussed in the subsequent sections.
 8.2 Advantages of Closed-Loop Control Systems 595

Besides Example 8.1.2, three additional examples of feedback control systems in applica-
tions are given in Section 8.7.

8.2 Advantages of Closed-Loop Control Systems

In this section, the three main advantages of closed-loop control as mentioned in the
previous section are illustrated in an example of motor speed control. Figure 8.2.1 shows
an open-loop control system and a closed-loop control system for the speed control of a field-
controlled DC motor. For simplicity in analysis, unity feedback without sensor noise has
been assumed for the closed-loop system. According to Section 6.4, a simplified model of the
motor (by neglecting the inductance of the field circuit) is given by

dω
τ þ ω ¼ Ko ð τ m þ τ d Þ (8.2.1)
dt

where ω is the motor rotation speed, τ and Ko are constants, τm is the motor torque, and τd is a
torque load applied to the motor shaft. Also, in the figure, ga is the gain constant of the
amplifier, and Ωr ðsÞ is a reference input in the s-domain, which is the desired motor speed.
Let the desired motor speed be ωr ¼ ω r , which is a prescribed constant. Let the torque
1
load be τd ¼ τ d ; which is a constant but can be unknown. Thus, Ωr ðsÞ ¼ ω r and
s
1
Τd ðsÞ ¼ τ d . According to the block diagrams in Figure 8.2.1, the output of the open-loop
s
system is given by

Ko 1
ΩðsÞ ¼ fga Co ðsÞω r þ τ d g (8.2.2)
τs þ 1 s

Td (s)
Controller Amplifier Plant
Ωr (s) U (s) Tm (s) + Ω (s)
+ Ko
Co(s) ga
τs + 1

(a)

Td (s) Plant
Controller Amplifier
+
Ωr (s) + E (s) U (s) Tm (s) + Ko Ω (s)
Ccl (s) ga
− τs + 1

(b)

Figure 8.2.1 Speed control of a DC motor: (a) open-loop control system; and (b) closed-loop control
system
596 8 Introduction to Feedback Control Systems

and the output of the closed-loop system is given by

Ko 1
ΩðsÞ ¼ fga Ccl ðsÞω r þ τ d g (8.2.3)
τs þ 1 þ ga Ccl ðsÞKo s

In Eqs. (8.2.2) and (8.2.3), Co ðsÞ and Ccl ðsÞ are the controller transfer functions to be
designed.

Advantage 1. Reduction of the Effects of Parameter Variations

In control engineering, the ability of a closed-loop system to reduce the effects of parameter
variations is also known as the robustness of the control system. For simplicity in discussion,
consider a zero load, τ d ¼ 0. Let the open-loop controller be Co ðsÞ ¼ Kc , where Kc is a
constant gain. The steady-state speed of the motor, from Eq. (8.2.2) and by the final-value
theorem (Section 2.4.3), is

Ko 1
ωss ¼ lim sΩðsÞ ¼ lim s ga Kc ω r ¼ Ko ga Kc ω r (8.2.4)
s→o s→o τs þ 1 s

Thus, if the control gain is chosen as Kc ¼ 1=ðKo ga Þ, the steady-state speed of the motor is the
same as the desired speed; that is ωss ¼ ω r .
Now we assume that the motor has parameter variations such that Ko become Ko þ ΔKo ,
with ΔKo being an unknown constant. Parameter variations such as ΔKo usually come from
uncertainties in system modeling or parameter drifting during operation. The steady-state
speed of the motor under open-loop control with Kc ¼ 1=ðKo ga Þ is

ΔKo
ωΔK
ss ¼ ðKo þ ΔKo Þga Kc ω r ¼ ω r þ
o
ωr (8.2.5)
Ko

Define the percentage error of the motor’s steady-state speed caused by the parameter
variations as
Δωss ωΔK o
 ωss
εΔKo ¼ ¼ ss (8.2.6)
ωss ωss

If follows from Eq. (8.2.5) that


ΔKo 1 ΔKo
εΔKo ¼ ωr þ ωr  ωr ¼ (8.2.7)
Ko ωr Ko

According to Eq. (8.2.7), the percentage error of the motor speed is directly related to the
percentage change of Ko , and the open-loop control has no influence on the parameter
 8.2 Advantages of Closed-Loop Control Systems 597

variations. For example, a 10% parameter variation ðΔKo =Ko ¼ 10%Þ) renders a 10% error in
the motor speed ðΔωss =ω r ¼ 10%Þ, which is not acceptable in application.
Under closed-loop control, if Ccl ðsÞ ¼ Kc (which describes P control as defined in Section
8.3), the steady-state speed of the motor, from Eq. (8.2.3) and using the final-value theorem,
is given by

Ko 1 Ko ga Kc
ωss ¼ lim s ga Kc ω r ¼ ωr (8.2.8)
s→o τs þ 1 þ ga Kc Ko s 1 þ Ko ga Kc

Under P control with a large enough control gain Kc , the motor speed can be close to the
desired speed. With the parameter uncertainty Ko þ ΔKo , the steady-state output ωΔK
ss of the
o

system is

ðKo þ ΔKo Þga Kc

ωΔK
ss ¼
o
ωr (8.2.9)
1 þ ðKo þ ΔKo Þga Kc

The percentage error of the motor speed, defined in Eq. (8.2.6), is obtained as

ðKo þ ΔKo Þga Kc ω r ð1 þ εÞKo ga Kc ω r

εΔKo ¼ 1¼ 1 (8.2.10)
1 þ ðKo þ ΔKo Þga Kc ωss 1 þ ð1 þ εÞKo ga Kc ωss

where ε ¼ ΔKo =Ko . For relatively small parameter variations ðε2 << 1Þ, the first term on the
right-hand side of Eq. (8.2.10) is reduced to

ð1 þ εÞKo ga Kc ω r Ko ga Kc 1 ð1 þ εÞ
¼ ωr
1 þ ð1 þ εÞKo ga Kc ωss 1 þ Ko ga Kc ωss εKo ga Kc
1þ
( 1 þ Ko ga Kc )
 2
1 Ko ga Kc K o ga K c
¼ ωss ð1 þ εÞ 1  ε þ ε2  ...
ωss 1 þ Ko ga Kc 1 þ Ko ga Kc

Ko ga Kc 1
≈1þε 1 ¼1þε
1 þ Ko g a K c 1 þ Ko ga Kc

1
where Eq. (8.2.8) and the series ¼ 1  x þ x2  x3 þ . . . (for jxj < 1) have been used,
1þx
and the ε2 -term and higher-order terms have been eliminated. It follows that

1 ΔKo
εΔKo ≈ (8.2.11)
1 þ Ko ga Kc Ko

This indicates that the percentage error of the motor speed can be significantly reduced by
properly assigning the control gain Kc . For instance, for a 10% parameter variation,
ΔKo =Ko ¼ 10%, the selection of Kc satisfying the condition 1 þ Ko ga Kc ≥ 20 results in
Δωss =ωss ≤ 0:5%, which is much smaller than that with open-loop control.
598 8 Introduction to Feedback Control Systems

1
Alternatively, let us now consider PI control (Section 8.3), Ccl ðsÞ ¼ KP þ KI . It is easy to
s
show that the steady-state motor speed without parameter variations is

ga Ko ðKP s þ KI Þ
ωss ¼ lim sΩðsÞ ¼ lim ωr
s→o s → o τs2 þ s þ ga Ko ðKP s þ KI Þ
g a Ko KI
¼ ωr ¼ ωr (8.2.12)
g a Ko KI

Here, it is easy to check that sΩðsÞ satisfies the condition of the final-value theorem because
all the parameters of the motor and controller are positive. Replacing Ko by Ko þ ΔKo in Eq.
(8.2.12) gives the steady-state output of the control system as follows:

ga ðKo þ ΔKo ÞKI

ωΔK
ss ¼
o
ωr ¼ ωr (8.2.13)
ga ðKo þ ΔKo ÞKI

Equations (8.2.12) and (8.2.13) show that, under PI control, the motor reaches the desired
speed, with or without the parameter variation ΔKo . Therefore, the percentage error of the
motor speed as defined in Eq. (8.2.6) is zero for any value of ΔKo ; namely,
ωr  ωr
εΔKo ¼ ¼ 0.
ωr

Advantage 2. Disturbance Rejection

We now compare open-loop control and closed-loop control in relation to a disturbance. To
this end, assume the motor is subject to a constant load τ d . For simplicity, we assume that the
motor has no parameter variation. For the open-loop system with the controller
Co ðsÞ ¼ Kc ¼ 1=ðKo ga Þ, its steady-state output by Eq. (8.2.2) is

Ko 1
ωτssd ¼ lim ga ω r þ τ d ¼ ω r þ Ko τ d (8.2.14)
s → o τs þ 1 Ko ga

Without disturbance, the steady-state output of the open-loop system from Eq. (8.2.4) is
ωss ¼ ω r . The percentage error of the motor’s steady-state speed caused by the disturbance
τ d is defined as

Δωss ωτssd  ωss

ετ d ¼ ¼ (8.2.15)
ωss ωss

From Eq. (8.2.14),

Δωss ω r þ Ko τ d  ω r Ko
ετ d ¼ ¼ ¼ τd (8.2.16)
ωss ωr ωr
 8.2 Advantages of Closed-Loop Control Systems 599

which indicates that the open-loop controller has no influence on the reduction of the output
error caused by the disturbance. Indeed, an increased τ d yields an increased error in the
motor speed.
Now, consider the closed-loop control system in Figure 8.2.1(b), with a disturbance τ d and
Ccl ðsÞ ¼ Kc (P control). The steady-state output of the control system, from Eqs. (8.2.3) and
(8.2.8) and by the final-value theorem, is

Ko
ωτssd ¼ lim fga Kc ω r þ τ d g
s → o τs þ 1 þ ga Kc Ko
ga Kc Ko Ko τ d Ko τ d
¼ ωr þ ¼ ωss þ (8.2.17)
1 þ ga Kc Ko 1 þ g a Kc Ko 1 þ ga Kc Ko

It follows from Eq. (8.2.15) that


Δωss Ko τ d 1 þ Ko ga Kc τ d 1
ετ d ¼ ¼ ωss þ  ωss ¼ (8.2.18)
ωss 1 þ ga Kc Ko K o g a Kc ω r ω r ga Kc

Thus, the percentage error of the motor’s steady-state speed decreases as the control gain Kc
increases.
1
If PI control is applied, Ccl ðsÞ ¼ KP þ KI , the steady-state speed of the motor with a
s
disturbance τ d is
 
τd Ko 1 1
ωss ¼ lim s  ga KP þ KI ω r þ τ d ¼ ω r
s→0 1 s s
τs þ 1 þ ga KP þ KI Ko
s

This means that the steady-state speed of the PI-controlled motor under the disturbance is
the same as the desired speed. Also, by setting τ d ¼ 0 in the previous equation, it can be
shown that the steady-state speed ωss of the motor without disturbance is also ω r . It follows
that the percentage error of the motor’s steady-state speed is zero; namely,

Δωss ω r  ω r
ετ d ¼ ¼ ¼0 (8.2.19)
ωss ωr

This example shows that closed-loop control can reject the effects of disturbances on the
control system output.

Advantage 3. Improvement of System Response

Closed-loop (feedback) control can improve the transient response, the steady-state
response, and the stability of a dynamic system. To show this, the motor parameters and
amplifier gain in Figure 8.2.1 are assigned as τ ¼ 0:5; Ko ¼ 3; ga ¼ 16 and the desired motor
600 8 Introduction to Feedback Control Systems

Open-Loop Step Response

100

90

80

70

Amplitude 60

50

40

30

20

10

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time (seconds)

Figure 8.2.2 The output of the open-loop control system in Figure 8.2.1(a) with Co ðsÞ ¼ Kc ¼ 1=48 and
τd ¼ 0

speed as ω r ¼ 100. Here, nondimensional values of the parameters have been considered.
Also, for simplicity, we assume that the motor does not have parameter variations, and is not
subject to a disturbance.
The output of the open-loop system in Figure 8.2.1(a), with Co ðsÞ ¼ 1=ðKo ga Þ ¼ 1=48, and
τ d ¼ 0 in Eq. (8.2.2), is given by

1 ωr 100 1
Ωo ðsÞ ¼ ¼ (8.2.20)
τs þ 1 s 0:5s þ 1 s

which by inverse Laplace transform gives the time-domain response

 
ωo ðtÞ ¼ 100 1  e2t (8.2.21)

The steady-state output of the open-loop system is ωo;ss ¼ lim ωo ðtÞ ¼ ω r ¼ 100. The output
s→o
of the open-loop system is plotted against time in Figure 8.2.2. We define a 2% settling time ts
as the time for the output to reach a 2% neighborhood of the final value ωo;ss . In other words,
ωo ðts Þ ¼ 98%  ωo;ss ; which from Eq. (8.2.21) gives ts ¼ 0:5lnð0:02Þ ¼ 1:956 s.
The above-defined settling time is a measure of the system performance. (Refer to Section
7.1 for system response analysis in the time domain.) A shorter settling time means that the
system response to the reference input ωr is faster. As can be seen from Eq. (8.2.20), the
settling time of the open-loop system is the same as that of the plant (motor). Thus, the open-
loop control gain Kc has no influence on the settling time.
 8.3 PID Control Algorithm 601

Closed-Loop Step Response

100

90

80

70

60
Amplitude
50

40

30

20

10

0
0 0.1 0.2 0.3 0.4 0.5
Time (seconds)

Figure 8.2.3 The output of the closed-loop control system in Figure 8.2.1(b) with Ccl ðsÞ ¼ 0:2 þ 0:4=s and
τd ¼ 0

To improve the system response with a shorter settling time, consider the closed-loop
1
control system in Figure 8.2.1(b), with a PI controller, Ccl ðsÞ ¼ KP þ KI . With careful
s
design, the controller parameters are chosen as KP ¼ 0:2; KI ¼ 0:4. The output of the
closed-loop system, by Eq. (8.2.3), is obtained as

ga Ko ðKP s þ KI Þ ωr 4800 ð0:2s þ 0:4Þ 1

Ωcl ðsÞ ¼ ¼ (8.2.22)
τs2 þ ð1 þ ga Ko KP Þs þ ga Ko KI s 0:5s2 þ 10:6s þ 48  0:4 s

By using the MATLAB function step, the output of the closed-loop system versus time is
plotted in Figure 8.2.3, which shows that the 2% settling time of the system is 0.204 s. Hence,
the motor speed under the closed-loop control reaches the desired value much faster than
that under the open-loop control.

8.3 PID Control Algorithm

8.3.1 PID Controller
The PID controller, which is also known as the three-term controller, is probably the most
commonly used control algorithm in engineering applications. The PID controller has the
transfer function form
602 8 Introduction to Feedback Control Systems

UðsÞ 1
Gc ðsÞ ¼ ¼ KP þ KI þ KD s (8.3.1)
EðsÞ s

where UðsÞ and EðsÞ are the controller output and error, as shown in Figure 8.1.3(b), and KP ,
KI , and KD are the control gains. It follows that the controller output in the time domain is

ðt
deðtÞ
uðtÞ ¼ KP eðtÞ þ KI eðτÞdτþKD (8.3.2)
dt
t0

The PID control algorithm given in Eq. (8.3.1) can also be written as

1
Gc ðsÞ ¼ KP 1 þ þ TD s (8.3.3)
TI s

where TI is the integral time and TD is the derivative time. Comparison of Eqs. (8.3.1) and
(8.3.3) shows that KI ¼ KP =TI and KD ¼ KP TD .
As can be seen from Eq. (8.3.2), the controller output has three terms, with the first term
Proportional to the error eðtÞ, the second term given by an Integral about the error, and the
third term related to a Derivative of the error. Thus, the PID control represents a combin-
ation of proportional, integral, and derivative control actions; see Figure 8.3.1, where the
three blocks describe the actions of P control, I control, and D control, respectively.
By setting certain control gains to zero, the PID controller is reduced to the following three
commonly used controllers.

(i) proportional (P) controller:

Gc ðsÞ ¼ KP (8.3.4)

Figure 8.3.1 A block

diagram of the PID
controller
 8.3 PID Control Algorithm 603

(ii) proportional plus integral (PI) controller:

1
Gc ðsÞ ¼ KP þ KI (8.3.5)
s

(iii) proportional plus derivative (PD) controller:

Gc ðsÞ ¼ KP þ KD s (8.3.6)

Thus, the PID, P, PI, and PD control algorithms are proper combinations of P, I, and D
control actions.

Example 8.3.1
In Figure 8.3.2, a first-order plant with parameters T ¼ 2 and b = 3 is under PID feedback
control. Consider a step reference input r(t) = 1. Assume a step disturbance d(t) = 2. Select
the control gain constants KP ; KI ; and KD such that the closed-loop system meets the following
control objectives:

(i) the system is stable

(ii) the steady-state error ðess ¼ lim eðtÞÞ is zero
t→∞
(iii) the settling time of the output (2% criterion) is less than 3 seconds
(iv) the maximum overshoot of the output is less than 10%

Solution
Write the reference input and disturbance as rðtÞ ¼ r0 ; dðtÞ ¼ d0 . From Eqs. (8.1.2) and (8.1.4),
the output of the control system is

D(s)
PID controller + Plant
R(s) + E(s) 1 U(s) + b
Y(s)
KP + KI + KDs
− s Ts + 1

Figure 8.3.2 A first-order system under PID feedback control

604 8 Introduction to Feedback Control Systems

1  b b
KD s2 þ KP s þ KI RðsÞ þ DðsÞ
YðsÞ ¼ s Ts þ 1 Ts þ 1
1  b
1 þ KD s2 þ KP s þ KI (a)
 s  r0 Ts þ 1
b KD s2 þ KP s þ KI þ bd0
¼ s
sðTs þ 1Þ þ bðKD s2 þ KP s þ KI Þ

This is a second-order system with the characteristic equation

1 þ 3KP 3KI
s2 þ sþ ¼0 (b)
2 þ 3KD 2 þ 3KD

According to Table 7.2.1, the control system is stable if and only if all the coefficients of the
characteristic equation are positive. Thus, the stability conditions are

ð2 þ 3KD Þ > 0; ð1 þ 3KP Þ > 0; KI > 0 (c)

These conditions are met if KP , KI , and KD are all positive.

Under the stability conditions, the poles of the function
 
3 KD s2 þ KP s þ KI þ 6s
sYðsÞ ¼ (d)
ð2 þ 3KD Þs2 þ ð1 þ 3KP Þs þ 3KI

which is obtained with T ¼ 2; b ¼ 3; r0 ¼ 1; and d0 ¼ 2, all have negative real parts. Thus, the
steady-state response of the closed-loop system, by the final-value theorem, is
 
3 KD s2 þ KP s þ KI þ 6s
yss ¼ lim sYðsÞ ¼ lim ¼1 (e)
s→0 s → 0 ð2 þ 3KD Þs2 þ ð1 þ 3KP Þs þ 3KI

This implies that the steady-state error ess ¼ limt → ∞ ðr  yÞ ¼ 1  yss ¼ 0.

Through comparison of the characteristic equation (a) with the standard form

s2 þ 2ξωn s þ ω2n ¼ 0

we obtain
1 þ 3KP
2ξωn ¼ (f)
2 þ 3KD

An estimation of the settling time, from Eq. (7.1.70), is ts ¼ 4=ξωn < 3, which by Eq. (f) gives the
condition

1 þ 3KP 8
> (g)
2 þ 3KD 3
 8.3 PID Control Algorithm 605

Note that the above estimation is only approximate because the closed-loop transfer function is
not in the standard form of a second-order system and because the system has two inputs
(reference and disturbance). Nevertheless, the formulas in Eq. (7.1.70) provide useful guidance
in control system design.
For numerical simulation, we choose KI ¼ 14 and KD ¼ 4, for which, condition (g) gives
KP > 12:11. Figure 8.3.3 shows the curves of the step response of the control system for
KP ¼ 12; 15; and 20, which are plotted by using MATLAB. It can be seen that

for KP ¼ 12; ts ¼ 2:16 s; MP ¼ 11%

for KP ¼ 15; ts ¼ 2:22 s; MP ¼ 9%
for KP ¼ 25; ts ¼ 3:02 s; MP ¼ 6%

where ts and MP are the settling time and maximum overshoot, respectively. Here, an increase in
the proportional gain KP decreases the maximum overshoot, but makes the settling time longer. It
follows that for KP ¼ 15; KI ¼ 14; and KD ¼ 4, all the four control objectives are satisfied.
The MATLAB commands for computing the step response in the case of KP ¼ 15, KI ¼ 14, and
KD ¼ 4 are as follows:
Kp = 15; Ki = 14; Kd = 4;
num = [3*Kd 3*Kp+6 3*Ki];
den = [2+3*Kd 1+3*Kp 3*Ki];
sys = tf(num, den)
step(sys)
where tf and step are MATLAB functions (see Appendix B).

1.15

Kp = 12
1.1 Kp = 15
Kp = 25

1.05
y(t)

0.95

0.9

0.85
0 1 2 3 4 5 6 7 8
t (seconds)

Figure 8.3.3 Step-response curves of the PID feedback control system, with KI ¼ 14; KD ¼ 4, and
KP ¼ 12; 15; 25
606 8 Introduction to Feedback Control Systems

The control system design presented in Example 8.3.1 is a trial and error approach, which
is just for demonstrative purposes. The design of a closed-loop system is more efficient and
systematic with the methods of analysis and design given in Sections 8.4 to 8.6. To show this,
the same control system is redesigned by the root locus method in Example 8.6.6.

8.3.2 PID Gain Tuning

The PID gains have pros and cons. Consider the step response of a system under PID
control. (Refer to Section 7.1.4 for the description of system step response.) The effects of
proportional, integral, and derivative gains are summarized in Table 8.3.1. In general, a
proportional control action is added to improve the rise time; an integral control action is
added to eliminate steady-state error; and a derivative control action is added to improve the
overshoot or stability margin. Note that the effects of these control gains interfere with each
other. Therefore, Table 8.3.1 only provides a reference for the initial design of a PID
controller. To assure the performance and stability of a closed-loop system with PID control,
fine tuning of the control gains is necessary.
Given the pros and cons of P, I, and D control actions, the gains of a PID controller need to
be properly tuned. The process for selection of PID gains is usually called PID tuning or simply
gain tuning. There are many methods for tuning PID gains, including manual PID tuning,
Ziegler–Nichols tuning, tuning by root locus, and tuning by the Nyquist stability criterion. In
the remainder of this section, manual tuning and a Ziegler–Nichols tuning method are
described. Additionally, PID tuning by the root locus method is introduced in Section 8.5.

Manual Tuning
Manual tuning is a trial and error approach. In gain tuning, Table 8.3.1 is used as a reference,
and attention is paid to both the performance and stability of the control system. The trial and
error selection of the control gains in Example 8.3.1 can be viewed as manual PID tuning.

Ziegler–Nichols Gain Tuning

In the early 1940s, John G. Ziegler and Nathaniel B. Nichols developed two heuristic
methods for tuning PID controllers: one based on the ultimate gain and ultimate period of
a closed-loop response, the other based on the reaction curve of an open-loop response. The

Table 8.3.1 Effects of PID gains on closed-loop responses

Rise time Overshoot Settling time Steady-state error

Increasing KP Decrease Increase Minor change Decrease

Increasing KI Decrease Increase Increase Eliminate
Increasing KD Minor change Decrease Decrease No influence
 8.3 PID Control Algorithm 607

purpose of these methods is to achieve a fast closed-loop response without excessive oscilla-
tions and with excellent disturbance rejection. Because gain tuning methods do not require
an accurate model for the plant in consideration, they are commonly used in various control
applications.
In this section, Ziegler and Nichols tuning based on the ultimate gain and ultimate period,
which is also known as the Ziegler–Nichols ultimate-cycle method, is introduced. The method
considers the step response of a closed-loop system, which can be obtained either numerically
or experimentally. (For Ziegler–Nichols tuning based on reaction curves, refer to the refer-
ences listed at the end of this chapter.)
The Ziegler–Nichols ultimate-cycle method takes the following three steps.

(1) Initially, set the integral and derivative gains to zero (KI ¼ 0 and KD ¼ 0) and consider
the response of the closed-loop system with P control and subject to a step input.
Mathematically, the system response can be expressed as

1 KP Gp ðsÞ r0
yðtÞ ¼ L (8.3.7)
1 þ KP Gp ðsÞ s

where Gp ðsÞ is the plant transfer function, r0 is the amplitude of the step input, and L 1
is the inverse Laplace transform operator. Here, without a loss of generality, a unity-
feedback control system shown in Figure 8.3.4 is considered.
(2) Starting from a small value, gradually increase the proportional gain KP until the
closed-loop response shows sustained oscillations with constant amplitude, in
which case the closed-loop system becomes marginally stable (see Section 7.2
for the stability definition). Here, we assume that the feedback system becomes
unstable at a higher gain. Under the circumstances, the period Tu of the oscilla-
tions is called the ultimate period; and the proportional gain is called the ultimate
gain, which is given by

Ku ¼ KP jat ultimate period (8.3.8)

Analytically, the ultimate gain Ku and ultimate period Tu satisfy the closed-loop char-
acteristic equation

1 þ Ku Gp ð jωu Þ ¼ 0 (8.3.9)
pffiffiffiffiffiffiffi
where j ¼ 1 and ωu ¼ 2π=Tu . The values of Tu and Ku can also be determined by the
root locus method (see Section 8.5.3 for instance).
(3) With the ultimate gain Ku and ultimate period Tu determined in Step 2, the gains of a PID
controller are given by Table 8.3.2, where P control, PI control, and PD control are also
considered.
608 8 Introduction to Feedback Control Systems

Table 8.3.2 PID tuning by the Ziegler–Nichols ultimate-cycle method

Control algorithm KP KI KD

P control 0:5Ku  
PI control 0:45Ku 0:54Ku =Tu 
PD control 0:8Ku  0:1Ku Tu
PID control 0:6Ku 1:2Ku =Tu 3Ku Tu =40

Example 8.3.2
Consider the unity-feedback system in Figure 8.3.4, where the plant transfer function is

2
Gp ðsÞ ¼ (a)
sðs þ 2Þðs þ 3Þ
and a PID controller is implemented. Determine the control gains KP , KI , and KD by the Ziegler–
Nichols ultimate-cycle method and plot the response of the control system subject a unit step
input, rðtÞ ¼ 1.
Solution
Follow the three steps of the ultimate-cycle method. By Eq. (8.3.7), the step response of the closed-
loop system with P control (KI ¼ 0 and KD ¼ 0) is computed at several values of the gain KP .
Figure 8.3.5 shows the step responses at KP = 12, 14, 15, and 16, which are obtained by the
MATLAB functions tf and step. As indicated by Figure 8.3.5(c), the ultimate period and the
ultimate gain are Tu ¼ 2:565 and Ku ¼ 15.
According to Table 8.3.2, the gains of the PID controller are obtained as follows

Ku 3Ku Tu
KP ¼ 0:6Ku ¼ 9; KI ¼ 1:2 ¼ 7:0173; KD ¼ ¼ 2:8857 (c)
Tu 40
The transfer function of the closed-loop system is

YðsÞ 2ðKD s2 þ KP s þ KI Þ
¼ 2 (d)
RðsÞ s ðs þ 2Þðs þ 3Þ þ 2ðKD s2 þ KP s þ KI Þ

Using Eqs. (c) and (d), the step response of the PID-controlled system is plotted by MATLAB in
Figure 8.3.6. Note that the step response has a large maximum overshoot. This is because the

Figure 8.3.4 A unity-feedback control system

 8.3 PID Control Algorithm 609

Kp = 12 Kp = 14
1.8 2
1.6 1.8

1.4 1.6
1.4
1.2
1.2
1

y(t)
y(t)

1
0.8
0.8
0.6
0.6
0.4 0.4
0.2 0.2
0 0
0 2 4 6 8 10 12 0 2 4 6 8 10 12
t (seconds) t (seconds)
(a) (b)

Kp = 15 Kp = 16
2 2.5
1.8
2
1.6
2.565 s
1.4
1.5
1.2
y(t)

y(t)

1 1
0.8
0.5
0.6
0.4
0
0.2
0 −0.5
0 2 4 6 8 10 12 0 2 4 6 8 10 12
t (seconds) t (seconds)
(c) (d)

Figure 8.3.5 Step response of the closed-loop system with a P controller for KP gains of (a) 12,
(b) 14, (c) 15, and (d) 16, plotted with MATLAB

Ziegler–Nichols tuning method is developed to provide the best disturbance-rejection perform-

ance rather than the best input response performance. To reduce the overshoot, the D control gain
can be increased, as indicated in Table 8.3.1. For instance, choosing KD ¼ 4:0 and keeping the
other gains the same as given in Eq. (c) yields the step response in Figure 8.3.7.
610 8 Introduction to Feedback Control Systems

Ziegler-Nichols Gain Tuning

1.6

1.4

1.2

1
y(t)

0.8

0.6

0.4

0.2

0
0 2 4 6 8 10 12
t (seconds)

Figure 8.3.6 Closed-loop step response by the Ziegler–Nichols PID tuning: KP ¼ 9, KI ¼ 7:0173,
and KD ¼ 2:8857

Ziegler-Nichols Gain Tuning with increased KD

1.5

1
y(t)

0.5

0
0 2 4 6 8 10 12
t (seconds)

Figure 8.3.7 Closed-loop step response by the Ziegler–Nichols PID tuning with increased KD : KP ¼ 9,
KI ¼ 7:0173, and KD ¼ 4
 8.4 Control System Analysis 611

8.3.3 System Responses by Simulink

In numerical simulation, the response of a feedback control system can be generated through
use of the MATLAB functions tf and step, as shown in Examples 8.3.1 and 8.3.2. The time
response of a feedback control system can also be conveniently computed by Simulink. For
instance, for the feedback control system in Example 8.3.2, a Simulink model is constructed
as shown in Figure 8.3.8, where the control gains in Eq. (c) of the example are assigned.
Running a simulation on the model yields the step response shown in Figure 8.3.6. Also,
changing the values of the control gains in the Simulink model and running another simula-
tion again, produces the same time response plot as shown in Figure 8.3.7.
Refer to Sections 6.5 and 6.6 and Appendix B for the basics of time-domain block
diagrams and Simulink models.

8.4 Control System Analysis

In the analysis and design of a closed-loop control system, the response, stability, and steady-
state error of the system are examined, and the system performance is evaluated. These issues
are briefly presented in this section. It is assumed that the control system in consideration is
modeled by the transfer function formulation given in Section 8.1.

8.4.1 Transient Response and Steady-State Response

For a feedback control system described by Eq. (8.1.2), its response in the time domain can
be obtained by inverse Laplace transform

yðtÞ ¼ yR ðtÞ þ yD ðtÞ þ yN ðtÞ (8.4.1)

d
9
Kp +
+ 2
+− 7.0173 1 + +
s s 3 + 5s 2 + 6s
r Ki Add y
+
Gp(s)
2.8857 du/dt Add1

Kd

Figure 8.3.8 A Simulink model for the feedback control system in Example 8.3.2
612 8 Introduction to Feedback Control Systems

where
ðt
1
yR ðtÞ ¼ L fTR ðsÞRðsÞg ¼ gR ðt  τÞrðτÞdτ
0

ðt
1
yD ðtÞ ¼ L fTD ðsÞDðsÞg ¼ gD ðt  τÞdðτÞdτ
0

ðt
1
yN ðtÞ ¼ L fTN ðsÞNðsÞg ¼ gN ðt  τÞnðτÞdτ (8.4.2)
0

with the impulse response functions given by

gR ðtÞ ¼ L1 TR ðsÞg; gD ðtÞ ¼ L 1 TD ðsÞg; gN ðtÞ ¼ L 1 fTN ðsÞg (8.4.3)

The yR ðtÞ, yD ðtÞ, and yN ðtÞ are the system responses caused by the reference input, disturb-
ance, and sensor noise, respectively. In a control system design, the controller Gc ðsÞ is such
that yD ðtÞ and yN ðtÞ are minimized and yR ðtÞ meets specified performance requirements.
The steady-state response of the system, if there is one, can be determined by

yss ¼ lim sYðsÞ ¼ lim sfTR ðsÞRðsÞ þ TD ðsÞDðsÞ þ TN ðsÞNðsÞg (8.4.4)

s→0 s→0

Here, it has been assumed that all the poles of sYðsÞ have negative real parts so that the final-
value theorem can be applied (see Sections 2.4 and Section 7.1). In a control system design,
the gain parameters of the controller are chosen such that yss reaches a desired value.

Example 8.4.1
Consider the feedback control system in Example 8.1.2, with the parameters having the values:
m ¼ 1; c ¼ 1; gs ¼ 1; ga ¼ 5; KP ¼ 8; KD ¼ 2 . The reference input and the external force are
both step functions, rðtÞ ¼ 2 and FðtÞ ¼ 4. The system output in the s-domain is

ga ðKD s þ KP ÞRðsÞ þ FðsÞ 10ðs þ 4Þ 2 1 4

YðsÞ ¼ ¼ 2 þ 2
ms þ ðc þ gs ga KD Þs þ gs ga KP s þ 11s þ 40 s s þ 11s þ 40 s
2

where on the right-hand side, the first term is the response caused by the reference input, and the
second term by the external force. The system output is plotted in Figure 8.4.1 with MATLAB.
For comparison purposes, the system output without the external force,
 8.4 Control System Analysis 613

2.5

1.5

Output 1

0.5 y
y_R

0
0 0.2 0.4 0.6 0.8 1 1.2
Time (seconds)

Figure 8.4.1 The time response of the feedback control system in Example 8.4.1


10ðs þ 4Þ 2
yR ðtÞ ¼ L 1
s2 þ 11s þ 40 s

is also plotted in the figure. It can be seen that, without the disturbance FðtÞ, the steady-state
response of the control system is the same as the desired value set up by the reference input,
lim yR ¼ 2. With the disturbance, however, the steady-state response of the closed-loop system,
t→∞
by Eq. (8.4.4), becomes

10ðs þ 4Þ 2 1 4
yss ¼ lim s 2 þ ¼ 2:1
s→0 s þ 11s þ 40 s s2 þ 11s þ 40 s

8.4.2 Stability Analysis

The stability of a feedback control system is essentially important to its functionality,
performance, and safe operation. Indeed, an unstable control system is not usable at all,
and in the worst-case scenario it leads to catastrophic consequences. Stability concepts and
stability analysis for dynamic systems are presented in Section 7.2, and the results therein can
be directly applied to control systems.
For a closed-loop control system as described by Eq. (8.1.2), the transfer functions TR ðsÞ,
TD ðsÞ, and TN ðsÞ given in Eq. (8.1.4) have the same denominator. This is also seen in Example
8.4.1. Thus, the characteristic equation for a closed-loop system is given by
614 8 Introduction to Feedback Control Systems

1 þ LðsÞ ¼ 1 þ Gs ðsÞGc ðsÞGa ðsÞGp ðsÞ ¼ 0 (8.4.5)

The roots of Eq. (8.4.5) are the poles of the closed-loop system.
There are several methods for stability analysis of closed-loop control systems, including
direct examination of pole locations, the Routh–Hurwitz stability criterion (Section 7.2.3),
the root locus method (Sections 8.5 and 8.6), and the Nyquist stability criterion. In the
following example, the Routh–Hurwitz criterion is applied.

Example 8.4.2
In Figure 8.4.2, the mass–damper system in Example 8.1.2 is controlled by a PI controller, which has
the transfer function given in Eq. (8.3.5). The characteristic equation of the closed-loop system, from
Eq. (8.4.5), is

1 1
1 þ gs KP þ KI ga 2 ¼0
s ms þ cs

or

ms3 þ cs2 þ gs ga KP s þ gs ga KI ¼ 0

Let the parameter values be: m ¼ 1; c ¼ 1; gs ¼ 1; and ga ¼ 5. The characteristic equation of

the closed-loop system becomes

s3 þ s2 þ 5KP s þ 5KI ¼ 0

According to the Routh–Hurwitz criterion (Table 7.2.1), the closed-loop control system is stable if
and only if KP > KI > 0.
Choose the control gains as KP ¼ 2 and KI ¼ 1, satisfying the stability conditions. Assume the
same inputs as in Example 8.4.1; namely, rðtÞ ¼ 2 and FðtÞ ¼ 4. The output of the closed-loop
system is

Figure 8.4.2 Mass–damper system under PI feedback control in Example 8.4.2

 8.4 Control System Analysis 615

1  
YðsÞ ¼  þms2
cs 1
KP þ KI ga RðsÞ þ FðsÞ
1 1 s
1 þ gs KP þ KI ga 2
s ms þ cs

1 1
¼ 3 14 þ 10
s þ s2 þ 10s þ 5 s

The output of the closed-loop in the time domain is plotted in Figure 8.4.3. It can be seen that,
under the influence of the external force, the output of the control system has a steady-state value
that is the same as that of the system with the reference input only ðyss ¼ 2Þ. However, the
transient response in Figure 8.4.3 has large overshoots and a much longer settling time, compared
to that shown in Figure 8.4.1.

2.5

2
y(t)

1.5

0.5

0
0 5 10 15 20 25
t (seconds)

Figure 8.4.3 The time response of the feedback control system in Example 8.4.2

8.4.3 Steady-State Error

In designing a closed-loop control system, the estimation and reduction of its steady-state
error is an important issue. For the closed-loop system shown in Figure 8.1.3(b), its error is
 
EðsÞ ¼ RðsÞ  Gs ðsÞ YðsÞ þ NðsÞ
616 8 Introduction to Feedback Control Systems

Figure 8.4.4 Error of the feedback control system shown in Figure 8.1.3

As a special case, when zero noise ðNðsÞ ¼ 0Þ and unity feedback ðGs ðsÞ ¼ 1Þ are considered,
EðsÞ ¼ RðsÞ  YðsÞ, which means that the error is the difference between the desired output
(the reference input) and the actual output.
To determine the error of the control system under the influence of the reference input,
disturbance, and sensor noise, the block diagram in Figure 8.1.3(b) is re-arranged into an
equivalent form in Figure 8.4.4, where EðsÞ is selected as the output. It follows that the
output (error) of the system is

EðsÞ ¼ WR ðsÞRðsÞ þ WD ðsÞDðsÞ þ WN ðsÞNðsÞ (8.4.6)

where the transfer functions are given by

1 Gs ðsÞGp ðsÞ Gs ðsÞ

WR ðsÞ ¼ ; WD ðsÞ ¼ ; WN ðsÞ ¼ (8.4.7)
1 þ LðsÞ 1 þ LðsÞ 1 þ LðsÞ

with the loop transfer function LðsÞ ¼ Gs ðsÞGc ðsÞGa ðsÞGp ðsÞ.
The steady-state error of the control system is estimated by the final-value theorem as
follows

ess ¼ lim eðtÞ ¼ lim sEðsÞ

t→∞ s→0
(8.4.8)
¼ lim sfWR ðsÞRðsÞ þ WD ðsÞDðsÞ þ WN ðsÞNðsÞg
s→0

Here, it has been assumed that all the poles of sEðsÞ have negative real parts (the condition of
the final-value theorem condition). If only the effect of the reference input on the error is
examined, the steady-state error of the control system ðDðsÞ ¼ NðsÞ ¼ 0Þ is

sRðsÞ
ess ¼ lim sEðsÞ ¼ lim (8.4.9)
s→0 s → 0 1 þ LðsÞ
 8.4 Control System Analysis 617

Steady-State Error Subject to a Step Input

Assume that the closed-loop system is stable, which means that all the roots of the charac-
teristic equation 1 þ LðsÞ ¼ 0 have negative real parts. For a step input rðtÞ ¼ r0 , it is easy to
show that sEðsÞ satisfies the condition of the final-value theorem. Thus, the steady-state error
of the control system can be written as
r0
ess ¼ lim (8.4.10)
s→0 1 þ LðsÞ

Here, lim f1 þ LðsÞg ≠ 0 because the system is stable.

s→0
If the loop transfer function does not contain a pole at the origin of the s-plane, Lð0Þ is
finite and, as a result,
r0
ess ¼ (8.4.11)
1 þ Lð0Þ

with 1 þ Lð0Þ ≠ 0. Thus, the steady-state error is finite. If the loop transfer function has a pole
1
at the origin, it can be written as LðsÞ ¼ PðsÞ, with Pð0Þ ≠ 0. By Eq. (8.4.10),
s
r0 s lims → 0 r0 s
ess ¼ lim ¼ ¼0 (8.4.12)
s → 0 s þ PðsÞ Pð0Þ

Moreover, if LðsÞ contains more than one pole at the origin, ess ¼ 0 provided that the closed-
loop system is stable.
Equation (8.4.11) indicates that P and PD controllers, which do not contain a pole at the
origin, cannot render a zero steady-state response although they can reduce it with a large
gain. On the other hand, Eq. (8.4.12) implies that PI and PID controllers, which introduce a
pole at the origin, may eliminate the steady-state error. These observations are useful for
control system design.

Example 8.4.3
In this example, we compare the steady-state errors of the control systems shown in Examples
8.1.2 and 8.4.2. For the system with a PD controller in Figure 8.1.4, its steady-state error is

1
RðsÞ þ gs FðsÞ ðms2 þ csÞRðsÞ þ gs FðsÞ
EðsÞ ¼ þ cs
ms2 ¼ 2
1 ms þ cs þ gs ðKP þ KD sÞga
1 þ gs ðKP þ KD sÞga 2
ms þ cs

For step inputs rðtÞ ¼ r0 and f ðtÞ ¼ f0 ,

618 8 Introduction to Feedback Control Systems

ðms2 þ csÞr0 þ gs f0
sEðsÞ ¼ (a)
ms2 þ ðc þ KD Þs þ gs ga KP

Assume that parameters m; c; gs ; ga ; KD ; KP all are positive. This, by Table 7.2.1, implies that the
poles of sEðsÞ have negative real parts, satisfying the condition of the final-value theorem. It
follows from Eq. (a) that the steady-state error of the control system is

f0
ess ¼ lim sEðsÞ ¼ (b)
s→0 ga KP

This result has two indications. First, under PD feedback control, the steady-state error is nonzero
and it is caused by the external force (disturbance). Second, an increase in ga KP reduces the steady-
4
state error. Furthermore, with the data given in Example 8.4.1, ess ¼ ¼ 0:1, which agrees
58
with the plots in Figure 8.4.1.
Now, consider the control system with a PI controller in Figure 8.4.2. Its error is

1
RðsÞ þ gs 2 FðsÞ ðms2 þ csÞsRðsÞ þ gs sFðsÞ
EðsÞ ¼  ms þ cs ¼ 3
1 1 ms þ cs2 þ gs ga KP s þ gs ga KI
1 þ gs KP þ KI ga 2
s ms þ cs

Under the step inputs rðtÞ ¼ r0 and f ðtÞ ¼ f0 ,

ðms2 þ csÞr0 þ gs f0
sEðsÞ ¼ s (c)
ms3 þ cs2 þ gs ga KP s þ gs ga KI

To satisfy the condition of the final-value theorem, all the roots of the denominate sEðsÞ must have
negative real parts. This requires that the closed-loop system be stable. Note that parameters
m; c; gs ; and ga are all positive. According to Table 7.2.1, the stability conditions for the control
gain constants KP and KI are

cKP > mKI > 0: (d)

Under the conditions (d), the steady-state error of the control system by Eq. (c) is zero:

ess ¼ lim sEðsÞ ¼ 0: (e)

s→0

Thus, with PI control, the steady-state error of the control system that is caused by the disturbance
f ðtÞ can be eliminated. This is shown in the response plot in Figure 8.4.3.
For comparison, the errors of the control systems in Examples 8.4.1 (Example 8.1.2) and 8.4.2
are obtained with the given data, and they are
 8.5 The Root Locus Method 619

2s2 þ 2s þ 4 2s2 þ 2s þ 4
EPD ðsÞ ¼ ; E PI ðsÞ ¼
sðs2 þ 11s þ 40Þ s3 þ s2 þ 10s þ 5

where EPD ðsÞ is the error of the PD control system in Example 8.4.1, and EPI ðsÞ is the error of the
PI control system in Example 8.4.2. By using MATLAB, the errors are plotted against time in
Figure 8.4.5. It can be seen that ðePD Þss ¼ 0:1 and ðePI Þss ¼ 0.

2
e_P_D(t)
1.5 e_P_I(t)

0.5
Error

−0.5

−1

−1.5
0 5 10 15 20 25
t (seconds)

Figure 8.4.5 Comparison of the errors of the feedback control systems in Examples 8.4.1 and 8.4.2

8.5 The Root Locus Method

The root locus method is a commonly used tool for the analysis and design of linear
time-invariant control systems. The root locus method, which was introduced by W. R.
Evans in 1948, is a graphic technique to analyze stability and performance of closed-
loop systems and to design feedback controllers. As shown in Section 7.2, the stability
and performance of a feedback control system can be described in terms of the location
of the roots of its characteristic equation in the complex plane (the s-plane). The root
locus method is based on the correlation between root locations and system response
(such as rise time, overshoot, and settling time). In this section, the basic concepts of
the root locus method are presented and two methods for sketching a root locus are
introduced. The utility of the method in control system design is demonstrated in
Section 8.6.
620 8 Introduction to Feedback Control Systems

8.5.1 Basic Concepts

The characteristic equation for the closed-loop system in Figure 8.1.3(b) is

1 þ LðsÞ ¼ 0 (8.5.1)

with the loop transfer function LðsÞ ¼ Gs ðsÞGc ðsÞGa ðsÞGp ðsÞ. Assume that LðsÞ contains a
parameter k, which is either a gain of the controller or a parameter of any other component
(plant, sensor, or actuator). The root locus of the system is a plot of the location of the roots
of the characteristic equation (8.5.1) in the complex plane as k varies from zero to infinity. In
other words, the root locus is a cluster of trajectories of the closed-loop poles in the s-plane as
k is varied through positive values.
As an example, consider a system with the closed-loop characteristic equation

s2 þ 2s þ k ¼ 0 (8.5.2)

with k being a nonnegative parameter. The characteristic roots are

s1 pffiffiffiffiffiffiffiffiffiffiffi
¼ 1  1  k (8.5.3)
s2

The roots are real and distinct if k < 1; they are repeated at 1 for k = 1; and they become
complex conjugates if k > 1. By choosing k = 0, 0.1, 0.2, 0.3, …, 2.0, the locations of the
roots, computed by Eq. (8.5.3), are plotted in the s-plane in Figure 8.5.1(a), where the root
locations for k = 0 are denoted by crosses ðÞ, and the root locations for k > 0 are denoted by
dots. Of course, the root locations for other positive values of k can be determined.
Connecting all the dots gives a root locus of the system as shown in Figure 8.5.1(b), where
the arrows indicate the direction of root movement as k increases. The root locus plot
provides a picture of the behavior of the closed-loop poles as parameter k is varied,
which is useful for stability analysis, dynamic response prediction, and control system
design.
In the root locus method, consider the characteristic equation of a feedback system in the
following form

bðsÞ
1þk ¼ 1 þ kPðsÞ ¼ 0 (8.5.4)
aðsÞ

where k is a parameter under investigation, and aðsÞ and bðsÞ are polynomials of s

aðsÞ ¼ an sn þ an1 sn1 þ . . . þ a1 s þ a0

(8.5.5)
bðsÞ ¼ bm sm þ bn1 sm1 þ . . . þ b1 s þ b0
 8.5 The Root Locus Method 621

0.5
Im

−0.5

−1

−2 −1.5 −1 −0.5 0 0.5

Re
(a) (b)

Figure 8.5.1 Plot of the root locus of s2 þ 2s þ k ¼ 0 versus parameter k: (a) dots – root locations at
discrete values of k; (b) root locus by connecting the dots in (a)

where an > 0, bm > 0, and n ≥ m. For the purpose of generating a root locus with respect to k,
PðsÞ ¼ bðsÞ=aðsÞ is called the open-loop transfer function. Accordingly, the roots of aðsÞ are
called the open-loop poles and the roots of bðsÞ are called the open-loop zeros.
Depending on the problem in consideration, the open-loop transfer function PðsÞ may or
may not have the same format as the loop transfer function LðsÞ in Eq. (8.5.1). For instance,
for the closed-loop system in Figure 8.5.2(a), the characteristic equation is

1
1þk ¼0 (8.5.6)
sðs þ 1Þðs þ 2Þ

In this case, LðsÞ ¼ kPðsÞ, without any open-loop zero and with three open-loop poles 0, −1,
and −2. As another example, for the closed-loop system in Figure 8.5.2(b), the characteristic
equation is

4ðs þ 3Þ
1 þ LðsÞ ¼ 0; with LðsÞ ¼
sðs þ kÞ

with LðsÞ being the loop transfer function. The characteristic equation is rewritten as

sðs þ kÞ þ 4ðs þ 3Þ ¼ s2 þ 4s þ 12 þ ks ¼ 0
622 8 Introduction to Feedback Control Systems

(a)

(b)

Figure 8.5.2 Two closed-loop systems with parameter k (see text for details)

It follows that an equivalent characteristic equation is

s
1 þ kPðsÞ ¼ 0; with PðsÞ ¼ (8.5.7)
s2 þ 4s þ 12

Therefore, the open-loop transfer function PðsÞ does not have the same format
pffiffiffi as LðsÞ. From
Eq. (8.5.7), there is one open-loop zero and two open-loop poles 2  j2 2.
Although Eqs. (8.5.1) and (8.5.4) may look different in format, they have the same roots,
which are the poles of the closed-loop system. Also, the characteristic equation (8.5.4) can be
written as an equivalent polynomial equation of degree n:

aðsÞ þ kbðsÞ ¼ 0 (8.5.8)

the roots of which, namely the closed-loop poles, are dependent on the value of k. Denote the
roots of Eq. (8.5.8) by ϕl ðkÞ; l ¼ 1; 2; . . . ; n, where ϕl ðkÞ are functions of k. Mathematically,
the root locus of the closed-loop system with respect to k is expressed by

ϕl ðkÞ; k : 0 → þ ∞; l ¼ 1; 2; . . . ; n (8.5.9)

8.5.2 Properties of the Root Locus

This section presents some properties of the root locus of a closed-loop system, which are
conventionally used to sketch the root locus for minimum calculation. As described in
Section 8.5.3, the root locus can be easily generated through use of the software
MATLAB. Nevertheless, the root locus properties are still useful in analysis and design of
feedback control systems, as shown in the examples in Section 8.6.
 8.5 The Root Locus Method 623

Magnitude and Angle Criteria

Let s be a point on the root locus at a value of k. The s and k meet the following criteria:

jaðsÞj
Magnitude criterion k¼ (8.5.10a)
jbðsÞj
Angle criterion ∠bðsÞ  ∠aðsÞ ¼ ð2l þ 1Þ180° (8.5.10b)

with l ¼ 0; 1; 2; 3; . . ., where aðsÞ and bðsÞ are from the characteristic equation (8.5.4).

Properties of the Root Locus

Assume that a feedback control system described by Eq. (8.5.4) has n open-loop poles
ðp1 ; p2 ; . . . ; pn Þ and m open-loop zeros ðz1 ; z2 ; . . . ; zm Þ, with n ≥ m. The open-loop poles
and zeros satisfy the polynomial equations

aðpi Þ ¼ 0; i ¼ 1; 2; . . . ; n
(8.5.11)
bðzj Þ ¼ 0; j ¼ 1; 2; . . . ; m

In a root locus plot, the crosses ðÞ denote open-loop pole locations and circles (°) denote
open-loop zero locations. The root locus defined in Section 8.5.1 has the following five
properties.

Property 1: Branches
The root locus has n branches, representing the movement of the n poles of the closed-loop
system, as shown in Eq. (8.5.9). Because the roots of Eq. (8.5.4) are either real or complex
conjugates, the branches are symmetric about the real axis in the s-plane.
According to Eq. (8.5.8), each branch starts at an open-loop pole. If n = m, all the branches
end at the open-loop zeros. If n > m, m branches end at the open-loop zeros and the
remaining n − m branches go to infinity in the s-plane as k → ∞. (The way in which these
n − m branches go to infinity is described in Property 3.) This is illustrated in Figure 8.5.3,
where the arrows indicate the direction of root movement as k increases.

Property 2: Real-Axis Segments

If there exist any real open-loop poles or zeros, the root locus has segments on the real axis,
which are located to the left of an odd number of the real-axis poles and zeros. Two cases of
real-axis segments of root loci are shown in Figure 8.5.4.

Property: 3 Asymptotes
In the case of n > m, the n − m branches going to infinity approach asymptotes as k → ∞.
These asymptotes are a cluster of radial lines with a real-axis intercept σa and n − m angles ϕa ,
which are given by
624 8 Introduction to Feedback Control Systems

(a) (b)

Figure 8.5.3 Branches of root loci: (a) n = m = 3; (a) n = 3 and m = 2, with one branch going to negative
infinity on the real axis

Figure 8.5.4 Two cases of real-axis segments of root loci

( )
1 Xn Xm
σa ¼ pi  zj (8.5.12)
n  m i¼1 j¼1

180°
ϕa ¼ ð2l  1Þ; l ¼ 1; 2; . . . ; n  m (8.5.13)
nm

As an example, consider a feedback control system with the characteristic equation

sþ2
1þk ¼ 0; k > 0 (8.5.14)
sðs þ 3Þðs2 þ 2s þ 5Þ

From Eq. (8.5.14), the open-loop poles and zero are found as follows

p1 ¼ 0; p2 ¼ 3; p3 ¼ 1 þ j2; p4 ¼ 1 þ j2; z1 ¼ 2 (8.5.15)

Because n – m = 4 – 1 = 3, the root locus has three asymptotes, with the real intercept and
angles given by
 8.5 The Root Locus Method 625

1
σA ¼ fð0  3  1 þ j2  1  j2Þ  ð2Þg ¼ 1
41
180°
ϕA ¼ ð2l  1Þ ¼ 60°; 180°; 300°; for l ¼ 1; 2; 3 (8.5.16)
41

Figure 8.5.5 shows the asymptotes of the root locus.

Table 8.5.1 lists the asymptote angles with respect to the values of n – m. As observed from
the table, the larger the n – m, the more asymptotes stretch into the right half-plane. This
indicates that a feedback system with a larger n − m value is more inclined to be unstable as k
increases. Accordingly, adding a pole to the open-loop transfer function PðsÞ in Eq. (8.5.4),
as in I control, can make the closed-loop system less stable and adding a zero, as in PD
control, can enhance the system stability.

Property 4: Breakaway and Break-in Points on the Real Axis

If a real-axis segment of the root locus is connected to two open-loop poles, there is a point on the
segment at which the root locus leaves the real axis and branches out into the complex region; see
Figure 8.5.6(a), where the arrows indicate the direction of root movement as k increases. This
point ðsb Þ is called the breakaway point. On the other hand, if a segment of the root locus is
connected to two open-loop zeros, there is a point on the segment at which the two branches
of the root locus return from the complex region to the real axis; see Figure 8.5.6(b). This

Im

j3
Asymptote
x j2

j1
Asymptote
σA 60°

−4 −3 −2 −1 −60° 0 1 Re

−j 1

x −j 2
Asymptote

−j 3

Figure 8.5.5 Asymptotes of the root locus described by Eq. (8.5.14)

Table 8.5.1 Asymptote angles versus n − m

n−m 1 2 3 4

ϕA 180° 90°; 270° 60°; 180°; 300° 45°; 135°; 225°; 315°
Asymptotes Im Im Im Im

σA σA σA σA

Re Re Re
Re
 8.5 The Root Locus Method 627

(a) (b)

Figure 8.5.6 (a) Breakaway or (b) break-in point ðsb Þ

point ðsb Þ is called the break-in point. Breakaway and break-in points provide important
information about the change in the pattern of system response.
The breakaway and/or break-in points are determined as follows. The characteristic
equation (8.5.4) is rewritten as

aðsÞ 1
k ¼ FðsÞ; with FðsÞ ¼  ¼ (8.5.17)
bðsÞ PðsÞ

A breakaway or break-in point, denoted by sb , satisfies the following two conditions

dk d
¼ Fðsb Þ ¼ 0 (8.5.18a)
ds ds

and
k ¼ Fðsb Þ > 0 (8.5.18b)

Property 5: Crossover Points

The root locus may cross the imaginary axis at certain points; see Figure 8.5.7, where the
arrows indicate the direction of root movement as k increases. The points where the root
locus passes through the imaginary axis are called the crossover points or the imaginary axis
crossings. From Eq. (8.5.8), a crossover point ðs ¼ jωÞ satisfies the equation
pffiffiffiffiffiffiffi
aðjωÞ þ kbðjωÞ ¼ 0; j¼ 1 (8.5.19)

where ω is called the crossover frequency. The solution of Eq. (8.5.19) for k > 0 gives the
crossover points.
Recall that the real parts of the closed-loop poles detect the system stability (Section 7.2).
Indeed, the left half of the s-plane defines a stable region. Therefore, a crossover point is a
point where the stability characteristics of the closed-loop system is about to change as k
increases (from being stable to unstable or vice versa). This is clearly seen in Figure 8.5.7.
628 8 Introduction to Feedback Control Systems

Im Im

jω jω

Re Re

(a) (b)

Figure 8.5.7 Crossover points: (a) from a stable region to an unstable region; (b) from an unstable region
to a stable region

Also, at the crossover point, the value of ω yields the frequency of the oscillation and the
value of k defines a bound of the stability region.
Besides the aforementioned properties, the angle of departure of a branch at an open-loop
pole and the angle of arrival of a branch at an open-loop zero can be determined analytically.
The reader may refer to the texts listed at the end of this chapter.

8.5.3 Sketching Root Locus

There are two methods for sketching a root locus: (a) using Properties 1 to 5 as guides; and
(b) using the MATLAB function rlocus. Both the methods are often used alternatively in
analysis and design of feedback control systems.
In the first method, a root locus is roughly sketched by taking the following steps.

(1) Determine the open-loop poles and zeros.

(2) Locate the segments of the root locus on the real axis.
(3) Draw the asymptotes of the root locus if n > m.
(4) Determine the breakaway and/or break-in points.
(5) Determine the crossover points.
(6) Complete the root locus sketch.

A root locus sketch can provide a quick preview in stability analysis and controller design, as
illustrated in Section 8.6.

Example 8.5.1
Plot the root locus for the characteristic equation (8.5.14) with respect to parameter k.
Solution
Follow the above-mentioned six steps. Steps 1 and 3 (open-loop poles, zeros, and asymptotes)
have been carried out previously, as shown in Eqs. (8.5.15) and (8.5.16). In Step 2, by Property 2,
 8.5 The Root Locus Method 629

two segments of the root locus on the real axis are identified: one between 0 and −2, the other
between −3 and −∞. In Step 4, no breakaway or break-in point is found.
In Step 5, according to Eq. (8.5.19), the crossover points are governed by
  pffiffiffiffiffiffiffi
jωð jω þ 3Þ ω2 þ j2ω þ 5 þ kð jω þ 2Þ ¼ 0; j¼ 1 (a)

Separating the real and imaginary parts of Eq. (a) gives two coupled equations

Real : ω4  11ω2 þ 2k ¼ 0 (b)

 
Imaginary : ω 5ω2 þ 15 þ k ¼ 0 (c)

From Eq. (c), one possible solution is ω = 0. However, with ω = 0, Eq. (b) yields k = 0, which
violates the condition (8.5.18b). So, ω ≠ 0 and Eq. (c) is reduced to

5ω2 þ 15 þ k ¼ 0 (d)
pffiffiffi
Equations (b) and (d) are solved to obtain ω ¼  6 ¼ 2:449 and k = 15. Thus, the crossover
points are j2:449. Finally, the root locus is sketched in Figure 8.5.8 in Step 6. As seen from the
root locus, the feedback control system becomes unstable for k > 15.

Figure 8.5.8 Root locus of the characteristic equation (8.5.14)

630 8 Introduction to Feedback Control Systems

Example 8.5.2
Consider a closed-loop system with the characteristic equation

2sðs þ 4Þ
1þ ¼ 0; k > 0 (a)
ðs þ 7Þðs2 þ 4s þ kÞ

Sketch the root locus of the system, with respect to parameter k.

Solution
Equation (a) is rewritten as

ðs þ 7Þðs2 þ 4s þ kÞ þ 2sðs þ 4Þ
¼ ðs þ 7Þsðs þ 4Þ þ 2sðs þ 4Þ þ kðs þ 7Þ ¼ 0

which leads to the following equivalent characteristic equation

sþ7
1þk ¼ 0; k > 0 (b)
sðs þ 4Þðs þ 9Þ

From Eq. (b), there are three open-loop poles at 0, −4, and −9 and one open-loop zero at −7. By
Property 2, the root locus has two real segments: one between 0 and −4, and the other between −7
and −9. According to Property 3, the root locus has two asymptotes (n – m = 2), with the real
intercept and angles given by

1
σa ¼ fð0  4  9Þ  ð7Þg ¼ 3
31
(c)
180°
ϕa ¼ ð2l  1Þ ¼ 90°; 270°; for l ¼ 1; 2
31

From Property 4 and Figure 8.5.6(a), there is a breakaway point between 0 and −4, which satisfies
the conditions (8.5.18a) and (8.5.18b). Thus,

d d sðs þ 4Þðs þ 9Þ
FðsÞ ¼ 
ds ds sþ7
(d)
ð3s2 þ 26s þ 36Þðs þ 7Þ  sðs2 þ 13s þ 36Þ
¼ ¼0
ðs þ 7Þ2

Because the breakaway point is in the region 0 < s < −4, s þ 7 ≠ 0. Thus, Eq. (d) is reduced to

ð3s2 þ 26s þ 36Þðs þ 7Þ  sðs2 þ 13s þ 36Þ ¼ 0

or
s3 þ 17s2 þ 91s þ 126 ¼ 0 (e)

Equation (e) has only one real root: −2.1187, at which FðsÞ ¼ 5:619 > 0. Therefore, the break-
away point is sb ¼ 2:1187, at which k ¼ 5:619.
 8.5 The Root Locus Method 631

To check if there is any crossover point, consider Eq. (8.5.19), which by Eq. (a) leads to two
coupled equations:
 
ω ω2 þ 36 þ k ¼ 0 (f)

13ω2  7k ¼ 0 (g)

Because ω ≠ 0 for k > 0, Eq. (f) is reduced to

ω2 ¼ 36 þ k (h)

Substituting Eq. (h) into Eq. (g) yields k ¼ 78, which contradicts the assumption that k > 0.
Thus, the root locus does not have any crossover point.
Finally, the root locus is sketched in Figure 8.5.9.

Figure 8.5.9 Root locus of the characteristic equation (a) of Example 8.5.2

In the second method, the MATLAB function rlocus is used to generate a root locus.
For the closed-loop characteristic equation (8.5.4), the MATLAB commands
num = [bm bm-1 . . . b1 b0];
den = [an an-1 . . . a1 a0];
sys = tf(num, den)
rlocus(sys)

yield the root locus. Here, ai and bj are the coefficients of the polynomials aðsÞ and bðsÞ given
in Eq. (8.5.5).
632 8 Introduction to Feedback Control Systems

Root Locus
10

Imaginary Axis (seconds−1)

4

−2

−4

−6

−8

−10
−12 −10 −8 −6 −4 −2 0 2 4 6
Real Axis (seconds−1)

Figure 8.5.10 Root locus of the system in Example 8.5.1

For instance, for the system in Example 8.5.1, the MATLAB commands
num = [1 2];
den = conv([1 3 0], [1 2 5]);
sys = tf(num,den)
rlocus(sys)

produce a root locus in Figure 8.5.10, which is comparable with the sketch in Figure 8.5.8.

8.6 Analysis and Design by the Root Locus Method

In this section, analysis and design of feedback control systems by the root locus method is
illustrated on several examples. The MATLAB function rlocus is used as a tool, together
with the basics of stability analysis (Section 7.2) and the properties of root loci (Section 8.5).

8.6.1 Stability Analysis

As mentioned in Section 7.2.2, the stability of a feedback control system is dependent upon
its pole locations. A system is stable if all its poles are in the left half-plane. A system is
unstable if any of its poles are in the right half-plane. A system is marginally stable if some
poles, which are not repeated, are on the imaginary axis and if all other poles are in the left
half-plane. With the above-described stability concept, the stability of a feedback control
 8.6 Analysis and Design by the Root Locus Method 633

system with respect to a parameter can be examined via the root locus of the system, and the
corresponding stability conditions can be determined.

Example 8.6.1
Consider the closed-loop system in Figure 8.5.2(a). By plotting the root locus of the system,
determine the range of k for the system to be stable.

Solution
For the closed-loop system, its characteristic equation (8.5.6) can be written as
1
1þk ¼0 (a)
s3 þ 3s2 þ 2s

By the MATLAB commands

num = [1];
den = [1 3 2 0];
sys = tf(num,den)
rlocus(sys)
the root locus is plotted in Figure 8.6.1. As seen from the figure, the root locus has three
asymptotes, indicating that the system becomes unstable if k is large enough.
To determine the stability region for k, in the MATLAB figure, left-click the branch of the locus
with the 60° asymptote (Table 8.5.1) at any point, which shows an information window about the
point, such as location and gain. Move the point along the branch by holding the mouse button

Root Locus
5

4 System: sys
Gain: 5.98
3 Pole: −0.000761 + 1.41i
Imaginary Axis (seconds−1)

Damping: 0.000539
2 Overshoot (%): 99.8
Frequency (rad/s): 1.41
1

−1

−2

−3

−4

−5
−6 −5 −4 −3 −2 −1 0 1 2 3 4
Real Axis (seconds−1)

Figure 8.6.1 Root locus of the closed-loop system in Example 8.6.1

634 8 Introduction to Feedback Control Systems

until it reaches the imaginary axis. At this point, the information window gives the approximate
location of the crossover point (jω = j1.41) and the gain (k = 5.98). From Properties 1 and 5 in
Section 8.5.2, the system is stable if 0 < k < 5:98. A more precise analysis by Eq. (8.5.19) or the
Routh–Hurwitz stability criterion (Section 7.2.3) yields the stability range as 0 < k < 6 and the
crossover point as j6.

Example 8.6.1 shows that the MATLAB function rlocus gives approximate numerical
results. Nevertheless, the function is efficient and convenient in analysis and design of general
feedback control systems.

Example 8.6.2
A feedback control system is shown in Figure 8.6.2, where an unstable plant is under P control.
Sketch the root locus of the feedback system and find the range of the control gain k for stability.
Solution
The closed-loop characteristic equation is

1
1þk ¼0 (a)
ðs2 þ 8s þ 25Þðs  1Þðs þ 4Þ

There are four open-loop poles at 4  j3, 1, and −4; there is no open-loop zero. With the
MATLAB commands
g1 = tf(1, [1 8 25]);
g2 = tf(1, [1 -1]);
g3 = tf(1, [1 4]);
sys = g1*g2*g3;
rlocus(sys)
the root locus of the feedback system is plotted in Figure 8.6.3. Furthermore, by clicking the
branches and dragging the points toward the imaginary axis, two crossover points are identified

Figure 8.6.2 A feedback control system in Example 8.6.2

 8.6 Analysis and Design by the Root Locus Method 635

by the information windows as follows: j1.97 at k = 260 and 0 at k = 100. By the root locus
properties, all the closed-loop poles lie in the left half-plane if k is larger than 100 and less than 260.
Therefore, the range of k for the stability of the closed-loop system is 100 < k < 260.

Root Locus
15

10 System: sys
Gain: 260
Imaginary Axis (seconds−1)

Pole: −0.00042 + 1.97i

Damping: 0.000213
5 Overshoot (%): 99.9
Frequency (rad/s): 1.97

0
System: sys
Gain: 100
Pole: −0.00906
−5 Damping: 1
Overshoot (%): 0
Frequency (rad/s): 0.00906

−10

−15
−15 −10 −5 0 5 10
−1
Real Axis (seconds )

Figure 8.6.3 Root locus of Example 8.6.2

8.6.2 System Response Analysis

Consider a stable system with multiple poles. The pole of the system closest to the imaginary
axis is called the dominant pole. The concept of dominant poles has been briefly discussed in
Section 7.1.6. For a system with one dominant pole that is real ðs ¼ σÞ, it may be approxi-
mated as a first-order system. For a feedback control system with two dominant poles that
are a complex conjugate pair ðs ¼ σ  jωÞ, it may be approximated as a second-order
system. With an approximate model of a dominant pole(s), the system can be designed by the
root locus method to meet specific system response requirements.

Example 8.6.3
Consider the feedback system in Figure 8.6.2. Determine the value of the gain k such that the
system is stable and that the system response subject to an input rðtÞ ¼ 1 meets the following two
requirements: (i) the settling time (5% criterion) is less than 8 s; and (ii) the maximum overshoot is
less than 5%. With the determined k value, plot the step response of the system. Also, discuss the
possibility for the control system to have a settling time less than 2 s.
636 8 Introduction to Feedback Control Systems

Solution
The root locus of the feedback system has been plotted in Figure 8.6.3, from which it can be seen
that, for 112 < k < 260, the system has two dominant poles in a complex conjugate pair. (The other
poles are also complex.) By Eq. (7.1.70), the settling time (5% criterion) for a standard second-
order system is ts ¼ 3=ξωn . Thus, we need to choose a point p on the branch with an asymptote of
angle 45°, at which ts < 8 s or

3
ξωn ¼ ReðpÞ > ¼ 0:375 (a)
8

Following the clicking and moving approach in Example 8.6.1, a zoomed-in root locus of the
system is plotted in Figure 8.6.4, where the information window shows

ξωn ¼ 0:569; at k ¼ 116 (b)

By Eq. (b), the estimated settling time is ts ¼ 3=0:569 ¼ 5:27 s, which meets condition (i). Note
that the system is stable at k = 116. The information window also shows that the estimated
overshoot is 1.06%, which satisfies condition (ii). Here, the word “estimated” is used because the k
value given in Eq. (b) is based on an approximated model of the dominant poles. These
conditions are yet to be validated by the step response of the feedback control system that has
four poles.

Root Locus
6

4
System: sys
Gain: 116
Imaginary Axis (seconds−1)

Pole: −0.569 + 0.393i

2 Damping: 0.823
Overshoot (%): 1.06
Frequency (rad/s): 0.691

−2

−4

−6
−6 −5 −4 −3 −2 −1 0 1 2 3
−1
Real Axis (seconds )

Figure 8.6.4 Root locus of Example 8.6.3: ξωn ¼ 0:568 at k ¼ 116

 8.6 Analysis and Design by the Root Locus Method 637

To plot the step response, the closed-loop transfer function is obtained as follows

YðsÞ kðs þ 4Þ
¼ 2 (c)
RðsÞ ðs þ 8s þ 25Þðs  1Þðs þ 4Þ þ k

where Figure 8.6.2 has been used. With the MATLAB commands
k = 116;
num = k*[1 4];
d1 = [1 8 25]; d2 = [1 3 -4];
den = conv(d1, d2)+ [0 0 0 0 k];
sys = tf(num, den)
step(sys)
the system step response is plotted in Figure 8.6.5.
The steady-state response of the system subject to a unit step input is

kðs þ 4Þ 1
yss ¼ lim s ¼ 29; for k ¼ 116
s→0 ðs2 þ 8s þ 25Þðs  1Þðs þ 4Þ þ k s

For a settling time of the 5% criterion, ts is determined by yðts Þ ¼ 0:95yss ¼ 27:55 or

yðts Þ ¼ 1:05yss ¼ 30:45. It follows from the information window in Figure 8.6.5 that the settling
time is ts ¼ 5:17 s. Also seen from the figure, the maximum amplitude of the response is
ymax ¼ 29:3, which leads to a maximum overshoot of 1.03%. Thus, the two requirements for the
step response are met with k = 116.
Step Response
30

System: sys
25 Time (seconds): 5.17
Amplitude: 27.6

20
y(t)

15

10

0
0 2 4 6 8 10
t (seconds)

Figure 8.6.5 Step response of the feedback system in Example 8.6.3

638 8 Introduction to Feedback Control Systems

For the feedback system to have a settling time less than 2 s, instead of condition (a), it needs to
satisfy the following condition

3
ξωn ¼ ReðpÞ > ¼ 1:5 (d)
2
However, as observed from the root locus in Figure 8.6.4, ξωn < 0:6 for the complex dominant
poles, indicating that condition (d) cannot be met for any k > 0. Therefore, the step response of the
feedback system with P control cannot have a settling time less than 2 s. In fact, with the
dominant-pole model, the shortest possible settling time is estimated as ts ¼ 3=0:5 ¼ 6 s. This
issue shall be revisited in Example 8.6.4.

8.6.3 Control System Design

The root locus method provides a useful and efficient tool for design of feedback control
systems. In this section, two control system design methods are introduced: (a) an aid to PID
gain tuning; and (b) control system design by zero placement.

Aid to PID Gain Tuning

In the Ziegler–Nichols ultimate-cycle method for PID gain tuning (Section 8.3.2), the
ultimate period Tu and ultimate gain Ku of a closed-loop system with P control need to be
determined first. According to Property 5 in Section 8.5.2, Tu and Ku are related to a
crossover point, as shown in Eq. (8.3.9). By the root locus method, these parameters can
be easily identified. For instance, for the unity-feedback system in Example 8.3.2, the root
locus of the characteristic equation

2
1 þ KP ¼0 (8.6.1)
sðs þ 2Þðs þ 3Þ

with respect to gain KP is plotted in Figure 8.6.6, from which the crossover frequency is
identified as ω = 2.45 at KP ¼ 15. It follows that Tu ¼ 2π=ω ¼ 2:565 and Ku ¼ KP ¼ 15. This
result is the same as that obtained by the step responses in Figure 8.3.5. The root locus
method avoids a trial-and-error process, as demonstrated in Example 8.3.2, in which a
sequence of step responses with different values of KP has to be computed.
The Ziegler–Nichols gain tuning by the root locus method involves two steps.
First, Tu and Ku are determined by a crossover point of the root locus for a
corresponding feedback system with P control. Second, the PID gains are determined
from Table 8.3.2.
 8.6 Analysis and Design by the Root Locus Method 639

Root Locus
5

4 System: sys
Gain: 15
Pole: −0.00138 + 2.45i

Imaginary Axis (seconds )

−1
Damping: 0.000566
3 Overshoot (%): 99.8
Frequency (rad/s): 2.45

−1
−4 −3 −2 −1 0 1
Real Axis (seconds−1)

Figure 8.6.6 Root locus of Eq. (8.6.1): crossover frequency ω = 2.45 at KP ¼ 15

Control System Design by Zero Placement

According to the root locus properties presented Section 8.5.2, placement of additional
open-loop zeros can significantly alter the pattern of the root locus of a closed-loop system,
so as to change the stability and response characteristics of the system. For instance, by
Properties 1 and 3, addition of a zero reduces the number of asymptotes by increasing the
number m of open-loop zeros, which in turn may improve the stability of a system. As
another example, by Properties 2 and 4, addition of a real zero may remove a breakaway
point from the real axis. Due to these features and others, zero placement has been widely
used in design of feedback systems.
To show how zero placement works, consider PID control. From Eq. (8.3.1), the transfer
function of a PID controller can be written as

1 ðs  z1 Þðs  z2 Þ
Gc ðsÞ ¼ KP þ KI þ KD s ¼ KD (8.6.2)
s s

with z1 þ z2 ¼ KP =KD and z1 z2 ¼ KI =KD . As indicated by Eq. (8.6.2), the PID controller
introduces one open-loop pole at the origin of the s-plane, and two open-loop zeros,
which can be either real ones or a pair of complex conjugates. Assuming that the PID
gains are all positive, it can be shown that the added zeros are all located in the left
half-plane.
640 8 Introduction to Feedback Control Systems

(a) (b) (c) (d)

Figure 8.6.7 Zero placement: (a) PID control with complex zeros; (b) PID control with real zeros; (c) PI
control; and (d) PD control

Also, a PI controller given in Eq. (8.3.5) can be expressed by

1 sz
Gc ðsÞ ¼ KP þ KI ¼ KP (8.6.3)
s s

with z ¼ KI =KP , and a PD controller given in Eq. (8.3.6) can be expressed by

Gc ðsÞ ¼ KP þ KD s ¼ KD ðs  zÞ (8.6.4)

with z ¼ KP =KD . For positive gains, the zero introduced by a PI or PD controller is
located on the negative real axis. Figure 8.6.7 illustrates zero placements by PID, PI,
and PD controllers. Besides these, other controllers, including lead compensators, lag
compensators, lead-lag compensators, and notch filters can also be designed by zero
placement
According to the previous discussion, the design of a feedback control system by zero
placement takes the following four steps.

(1) Properly locate the zero(s) of the controller in the s-plane. The zero placement here is
either an initial guess of the controller zero(s) at the beginning of the control system
design or in an iterative step that is requested in Step 2 or Step 4.
(2) Perform a root locus analysis via the MATLAB function rlocus. By moving
points on the root locus branches, check if the control objectives, which are the
specifications on the system stability and performance, are satisfied in terms of the
dominant pole(s). It they are, go to Step 3. If not, go back to Step 1 and adjust the
zero location(s).
(3) With the selected gain k in the root locus equation (8.5.4), which is obtained in Step 2,
compute the controller gains, such as KP ; KI ; and KD .
 8.6 Analysis and Design by the Root Locus Method 641

(4) With the controller gains obtained in Step 3, simulate the response of the closed-loop
system to validate the control objectives. If all the specifications are met, the control
system design is completed. If not, go back to Step 1 and relocate the zero(s).

Example 8.6.4
As found in Example 8.6.3, the P control for the feedback system cannot yield a step response with
a settling time (5% criterion) less than 2 s. In this example, we show that PD control can resolve
this issue. To this end, the P controller in Figure 8.6.2 is replaced by a PD controller as described
by Eq. (8.6.3). The characteristic equation of the feedback system then is
sz
1 þ KD ¼0 (a)
ðs2 þ 8s þ 25Þðs  1Þðs þ 4Þ

where z is the open-loop pole introduced by the controller. The control objectives are stated as
follows. Under a unit step input, the response of the feedback system has: (i) a settling time (5%
criterion) less than 2 s; and (ii) a maximum overshoot less than 5%.
By examining the root locus in Figure 8.6.4, one realizes that the two branches from the
breakaway point (near −0.6) are too close to the imaginary axis. Due to ξω < 0:6 in the P control,
further reduction of the settling time below 2 s is impossible. Based on this observation, place the
controller zero z at −3.75, in order to remove the breakaway point and to reduce the number of
asymptotes. Consequently, the original branches from the breakaway point disappear and the
branches starting at the complex poles 4  j3 stretch into the right half-plane by following the
60° asymptotes. This way, the two dominant poles of the system now are on the new branches
and they can be located further away from the imaginary axis, so as to increase the ξωn value.
To show the effect of the above-mentioned zero placement ðz ¼ 3:75Þ, the root locus of the
feedback system is plotted in Figure 8.6.8. By moving a point on the branch starting at the pole
4 þ j3, the complex dominant poles are selected as 1:89  j1:69 at KD ¼ 46. Based on the
approximate model of the dominant poles, the settling time is estimated as

3 3
ts ¼ ¼ ¼ 1:59 s
ξωn 1:89

where Eq. (7.1.70) has been used. Also, the maximum overshoot (from the information window) is
found as 2.98%. The gains of the PD controller from Eq. (8.6.4) are KP ¼ 172:5 and KD ¼ 46.
With the selected controller gains, the closed-loop transfer function is obtained as

YðsÞ ð172:5 þ 46sÞðs þ 4Þ

¼ (b)
RðsÞ ðs2 þ 8s þ 25Þðs  1Þðs þ 4Þ þ ð172:5 þ 46sÞ

by which the system step response is plotted in Figure 8.6.9 and the corresponding steady-state
response is found as yss ¼ 9:52. From the figure, the settling time is 1.47 s, and the maximum
642 8 Introduction to Feedback Control Systems

amplitude is ymax ¼ 9:62, which gives the maximum overshoot as 1.05%. Thus, with the PD
controller Gc ðsÞ ¼ 172:5 þ 46s, the control objectives are satisfied.

Root Locus
8

6 System: sys
Gain: 46
Pole: −1.89 + 1.69i
4
Imaginary Axis (seconds−1)

Damping: 0.745
Overshoot (%): 2.98
2 Frequency (rad/s): 2.53

−2

−4

−6

−8
−8 −7 −6 −5 −4 −3 −2 −1 0 1 2
−1
Real Axis (seconds )

Figure 8.6.8 Root locus of the feedback system with PD control in Example 8.6.4

Step Response
10

9
System: sys yss = 9.52
8 Time (seconds): 1.47
Amplitude: 9.04
7

6
y(t)

0
0 1 2 3 4 5 6
t (seconds)

Figure 8.6.9 Step response of the feedback system in Example 8.6.4

 8.6 Analysis and Design by the Root Locus Method 643

Example 8.6.5
For the feedback control system in Figure 8.6.10, design a controller Gc ðsÞ by the root locus
method such that, under a unit step input rðtÞ ¼ 1, the system response satisfies the following
control objectives:

(i) the steady-state error ess ¼ 0

(ii) the settling time (5% criterion) ts ≤ 2 s
(iii) the maximum overshoot Mp falls in the region 5% ≤ Mp ≤ 15%

Solution
Consider a PID controller given in Eq. (8.6.2). The closed-loop characteristic equation then is

ðs  z1 Þðs  z2 Þ
1 þ LðsÞ ¼ 1 þ KD ¼0 (a)
sðs þ 1Þðs þ 2Þ

where the open-loop poles are 0, −1, and −2 and the open-loop zeros are z1 and z2 that are to be
located. Let z1 and z2 be a pair of complex conjugates, as shown in Figure 8.6.7(a).
With the root locus properties, a rough root locus of the closed-loop system versus KD is
sketched in Figure 8.6.11(a), where the branch starting at pole −2 lies on the negative real axis
and moves to ∞; the other two branches split at a breakaway point between 0 and −1 and they
eventually end at the zeros z1 and z2 . As can be seen from the root locus, the control system is
stable for any KD > 0. Because the loop transfer function LðsÞ contains a pole at the origin (s = 0),
from Eq. (8.4.12), the steady-state error ess is zero if the closed-loop system is stable. Thus,
objective (i) is automatically met for any KD > 0.
Figure 8.6.11(a) also shows that, for a large enough KD , the closed-loop system has two
dominant poles p1 and p2 in a complex conjugate pair, and one real pole p3 that is further away
from the imaginary axis. Therefore, in the control system design, the zeros z1 and z2 are to be
properly placed such that the dominant poles meet objectives (ii) and (iii). To this end, a design
region for the dominant poles is defined as follows. Write
qffiffiffiffiffiffiffiffiffiffiffiffiffi
p1 ¼ ξω þ j 1  ξ 2 ω (b)

For objective (ii), ts ¼ 3=ξω ≤ 2 s, which by Eq. (7.1.70) and Eq. (b) yields

Figure 8.6.10 The feedback control system in Example 8.6.5

644 8 Introduction to Feedback Control Systems

3
ξω ≥ ¼ 1:5 (c)
2
From Eq. (7.1.66), objective (iii) is converted to

0:4559 ≤ ξ ≤ 0:6901 (d)

Then, by Eq. (7.1.59), condition (d) is equivalent to 46:46° ≤ θ ≤ 58:87°, where θ is the angle made
by p1 with the negative real axis; that is, θ ¼ π  ∠p1 (Figure 7.1.13). With the conditions (c) and
(d), a design region for the dominant pole p1 is drawn in Figure 8.6.11(b). A design region for p2 is
not necessary because the dominant poles are complex conjugates. Thus, in the control system
design, one only needs to check if the conditions (c) and (d) are satisfied.
With the design region, the zeros of the PID controller are chosen as
pffiffiffiffiffiffiffi
z1 ¼ 2 þ j2; z1 ¼ 2  j2; j ¼ 1 (e)

by which the characteristic equation (a) becomes

s2 þ 4s þ 8
1 þ KD ¼0 (f)
sðs þ 1Þðs þ 2Þ

Here, the zeros z1 and z2 , which are the end points of the root locus, are selected to guide the
branches through the design region. With Eq. (f), the root locus of the system versus KD is plotted
by MATLAB in Figure 8.6.12. The information window in the figure shows that at KD ¼ 10,
Reðp1 Þ ¼ 1:77, and ξ ¼ 0:61, satisfying conditions (c) and (d). In other words, with the selected
zeros and control gain, the closed-loop pole p1 falls in the design region. Thus, the designed PID
controller is given by
Im Im
p1
Design
z1 region

p1

p3

−2 −1 0 Re
θ2
θ1

− σ0 0
z2 Re

p2

(a) (b)

Figure 8.6.11 Design of the PID controller in Example 8.6.5: (a) a sketch of the root locus; (b) a
design region (gray area) for the dominant pole p1 , where σ0 ¼ 1:5, θ1 ¼ 46:36° and θ2 ¼ 58:87°
 8.6 Analysis and Design by the Root Locus Method 645

s2 þ 4s þ 8 1
Gc ðsÞ ¼ 10 ¼ 40 þ 80 þ 10s (g)
s s

Finally, to validate the designed control system, the closed-loop transfer function is obtained as
follows

YðsÞ 10ðs2 þ 4s þ 8Þ
¼ 3 (h)
RðsÞ s þ 13s2 þ 42s þ 80

By Eq. (h), the response of the closed-loop system subject to a unit step input is plotted in Figure
8.6.13 and the corresponding steady-state response is found as yss ¼ 1. From the step-response
plot, it is found that the settling time is 0.91 s and the maximum overshoot is 10%. This verifies that
the designed PID controller in Eq. (h) meets all the control objectives.

Root Locus
3

2
Imaginary Axis (seconds−1)

0
System: sys
Gain: 10
−1 Pole: −1.78 − 2.3i
Damping: 0.61
Overshoot (%): 8.89
Frequency (rad/s): 2.91
−2

−3
−3 −2.5 −2 −1.5 −1 −0.5 0 0.5

Real Axis (seconds−1)

Figure 8.6.12 Root locus of the feedback system with PID control in Example 8.6.5
646 8 Introduction to Feedback Control Systems

Step Response
1.2

0.8
y(t)

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3 3.5 4
t (seconds)

Figure 8.6.13 Step response of the feedback system in Example 8.6.5

Example 8.6.6
Redesign the control system in Example 8.3.1 by the root locus method to meet the same control
objectives.
Solution
According to Figure 8.3.2, the characteristic equation of the closed-loop system is

1:5ðs  z1 Þðs  z2 Þ
1 þ LðsÞ ¼ 1 þ KD ¼0 (a)
sðs þ 0:5Þ

where z1 and z2 are the zeros of the PID controller, and KD is the gain constant in Eq. (8.6.2).
Equation (a) indicates that the root locus has two branches that start at poles 0 and −0.5 and end
at zeros z1 and z2 . Assume complex controller zeros. A rough sketch of the root locus is shown in
Figure 8.6.14(a), where p1 and p1 are the complex closed-loop poles at a certain value of KD to be
determined. The figure shows that the closed-loop system is stable for any KD > 0. Because the
loop transfer function LðsÞ contains a pole at the origin (s = 0), the steady-state error ess by Eq.
(8.4.12) is zero if the closed-loop system is stable. Accordingly, the control objectives (i) and (ii) are
met for any positive Kp D . ffiffiffiffiffiffiffiffiffiffiffiffiffi
Write p1 ¼ ξω þ j 1  ξ 2 ω. Then, from Eq. (7.1.70), objective (iii), ts < 3 s (2% criterion), is
equivalent to

4
ξωn > ¼ 1:3333 (b)
3

From Eqs. (7.1.59) and (7.1.66), objective (iv), Mp < 10%, can be written as
 8.6 Analysis and Design by the Root Locus Method 647

ξ > 0:5912 or θ < 53:76° (c)

where θ is the angle made by p1 with the negative real axis, as shown Figure 8.6.14(a). By the
conditions (b) and (c), select the controller zeros as z1;2 ¼ 3  j1. The root locus with the
assigned zeros is plotted by MATLAB in Figure 8.6.14(b), in which the information window
shows that at KD ¼ 3:5 and the closed-loop pole p1 ¼ 2:56 þ j1:36. This implies that ξω ¼ 2:56
and ξ ¼ 0:883, satisfying the conditions (b) and (c).
It follows that the designed PID controller is

UðsÞ ðs  z1 Þðs  z2 Þ 1
¼ KD ¼ 18:07 þ 29:66 þ 3:53s (d)
EðsÞ s s

showing that KP ¼ 18:07; KI ¼ 29:66; and KD ¼ 3:53: By Eq. (a) in Example 8.3.1, the step
response of the PID control system is plotted in Figure 8.6.15. As shown in the figure, the system
response has a maximum overshoot of 8%, a settling time of 1.38 s, and zero steady-state error
ðess ¼ lim ðr  yÞ ¼ 0Þ. Thus, all the control objectives are satisfied by the designed controller.
t→∞
Compared with the controller design demonstrated in Example 8.3.1, the zero placement by the
root locus method presented herein is systematic and avoids trial and error.

Root Locus
2

1.5
System: sys
1 Gain: 3.5
Pole: −2.56 + 1.36i
Imaginary Axis (seconds−1)

Im
Damping: 0.883
p1 0.5 Overshoot (%): 0.269
Frequency (rad/s): 2.9
z1
0

−0.5
−σ θ
0 Re −1
−0.5

−1.5

z2
−2
p2 −3.5 −3 −2.5 −2 −1.5 −1 −0.5 0 0.5
Real Axis (seconds−1)

(a) (b)

Figure 8.6.14 Root locus of the feedback system in Example 8.6.6: (a) a rough sketch, σ ¼ ξωn
and cos θ ¼ ξ; (b) a locus for the controller design
648 8 Introduction to Feedback Control Systems

1.1

1.05

y(t)
0.95

0.9

0.85

0.8
0 0.5 1 1.5 2 2.5 3
t (seconds)

Figure 8.6.15 Step response of the PID feedback control system in Example 8.6.6, with KP ¼ 18:07,
KI ¼ 29:66, and KD ¼ 3:53

8.7 Additional Examples of Control Systems

In this section, three examples of feedback systems from control applications are presented,
with the focus on transfer function formulation and block diagram representation. To
analyze and design these control systems, the methods presented in Sections 7.1–7.2 and
Sections 8.4–8.6 may be applied.

Example 8.7.1 Robotic System

Figure 8.7.1 shows is a single-arm robotic system, which consists of a rigid arm, a sensor, a
feedback controller, an amplifier, and a field-controlled direct-current (DC) motor as an actuator.
The sensor measures the rotation angle θ of the arm, ys is the sensor output, r is the reference input,
e is the error signal, u is the controller output (control command), and v is the amplifier output. Let
the moment of inertia of the arm–motor-shaft assembly be IO , and the coefficient of the motor
bearing be b. The sensor equation is ys ¼ ks θ, where ks is a gain constant. The amplifier equation is
v ¼ ka u, where ka is a constant. Consider PD control. Let the system input and output be r and θ,
respectively.

(a) Draw a block diagram of the robotic system.

ΘðsÞ
(b) With the block diagram obtained in (a), determine the transfer function .
RðsÞ
 8.7 Additional Examples of Control Systems 649

Solution
The s-domain governing equations of the components of the robotic system are given as follows:

 
ðiÞ Arm : IO s2 þ bs ΘðsÞ ¼ Tm ðsÞ (a)

where Tm ðsÞ is the motor torque

 
ðiiÞ Motor : Lf s þ Rf If ðsÞ ¼ VðsÞ (b)
Tm ðsÞ ¼ KIf ðsÞ (c)

where the parameters are from Eqs. (6.4.25) and (6.4.27)

ðiiiÞ Amplifier : VðsÞ ¼ ka UðsÞ (d)

UðsÞ
ðivÞ Controller : ¼ KP þ KD s (e)
EðsÞ

where KP and KD are constant control gains

ðvÞ Error : EðsÞ ¼ RðsÞ  Ys ðsÞ (f)

ðviÞ Sensor : Ys ðsÞ ¼ ks ΘðsÞ (g)

Based on these equations and by following the block diagram rules in Section 6.2.3, the block
diagram of the robotic system is constructed in Figure 8.7.2. From the block diagram, the transfer
function of the system is obtained as follows:

ΘðsÞ ka KðKP þ KD sÞ
¼ (h)
RðsÞ sðIO s þ bÞðLf s þ Rf Þ þ ks ka KðKP þ KD sÞ

The robotic system is a third-order system.

Arm
Io
Sensor
θ

Motor
ys

v
−
u e + r
Amplifier Controller

Figure 8.7.1 Schematic of the robotic control system in Example 8.7.1

650 8 Introduction to Feedback Control Systems

Figure 8.7.2 Block diagram of the robotic control system in Example 8.7.1

Example 8.7.2 Thermal Control System

Figure 8.7.3 shows a thermal control system regulating the temperature of a container (e.g., a
thermal bottle). The operation of the control system is explained as follows. A thermometer
(sensor) measures the container temperature T; a controller with a control algorithm determines
its output u (control command) based on the error signal e, which is the difference between the
desired temperature Tr (reference input) and the sensor output Ts ; under the control command, the
heater (actuator) inputs a heat flow rate qc into the container, such that the temperature of
the container, under the influence of ambient temperature Ta , is maintained as desired.
Let the thermometer and heater be described by

Figure 8.7.3 Thermal control system in Example 8.7.2

 8.7 Additional Examples of Control Systems 651

T^ s ðsÞ ¼ gs T^ ðsÞ;
ga (a)
^q c ðsÞ ¼ ^
u ðsÞ
τa s þ 1

where ^a stands for the Laplace transform of a with respect to time, gs and ga are gain constants,
and τa is a time constant. As shown in Eq. (a), the heater is modeled as a first-order dynamic
system. Assume that PID control is adopted:

1
^u ðsÞ ¼ KP þ KI þ KD s ^e ðsÞ (b)
s
T^ðsÞ T^ðsÞ
where KP ; KI ; and KD are control gains. Derive the transfer functions and .
^
T r ðsÞ ^
T a ðsÞ
Solution
The dynamic equation of the container is

dT
C ¼ qc  qo (c)
dt

where C is the thermal capacitance of the container, and the conductive heat transfer through the
container wall with thermal resistance R is described by
1
qo ¼ ðT  Ta Þ (d)
R
The error signal in the s-domain is written as

^e ðsÞ ¼ T^ r ðsÞ  T^ s ðsÞ (e)

Substituting Eq. (d) into Eq. (c) yields

dT 1
C ¼ qc  ðT  Ta Þ
dt R

which, after Laplace transformation, becomes

q c ðsÞ þ T^ a ðsÞ
ðRCs þ 1ÞT^ ðsÞ ¼ R^ (f)

After substituting Eqs. (a), (b), and (e) into Eq. (f), we arrive at

Rga
ðRCs þ 1ÞT^ ðsÞ ¼ ^u ðsÞ þ T^ a ðsÞ
τa s þ 1 
Rga 1  
¼ KP þ KI þ KD s T^ r ðsÞ  gs T^ ðsÞ þ T^ a ðsÞ
τa s þ 1 s
652 8 Introduction to Feedback Control Systems

or
 
sðRCs þ 1Þðτa s þ 1Þ þ Rga gs KP s þ KI þ KD s2 gT^ ðsÞ
 
¼ Rga KP s þ KI þ KD s2 T^ r ðsÞ þ sðτa s þ 1ÞT^ a ðsÞ (g)

It follows from Eq. (g) that the transfer functions of the control system are given by

 
T^ðsÞ

Rga KP s þ KI þ KD s2
¼ (h)
T^r ðsÞ
T^ a ðsÞ¼0 sðRCs þ 1Þðτa s þ 1Þ þ Rga gs ðKP s þ KI þ KD s2 Þ

and


T^ðsÞ

sðτa s þ 1Þ

¼ (i)
^
T a ðsÞ T^r ðsÞ¼0 sðRCs þ 1Þðτa s þ 1Þ þ Rga gs ðKP s þ KI þ KD s2 Þ

Note that the two transfer functions have the same denominator.

Example 8.7.3 Liquid-Level Control System

A liquid-level control system is shown in Figure 8.7.4, where the liquid height h of a tank of
constant area A is measured by a sensor; a controller determines the control command u based on
the error e ¼ r  ys ; and with u, a valve actuator adjusts the input volume flow rate qin such that
the tank liquid height is maintained at a desired level specified by a reference input r. The sensor
and actuator outputs are given by ys ¼ μs h and qin ¼ μa u, respectively, with μs and μa being
constants. The controller is described by UðsÞ ¼ Gc ðsÞEðsÞ, where the transfer function Gc ðsÞ

Figure 8.7.4 The liquid-level control system in Example 8.7.3

 Chapter Summary 653

describes a control algorithm. In working principle, this control system is similar to a toilet tank,
in which an assembly of a float cap, a rod, and a fill valve has the functionality of the senor,
controller, and valve actuator, respectively. The only difference is that when the toilet tank is
storing water, the valve at its bottom is closed, qout ¼ 0.
HðsÞ
Consider P control: Gc ðsÞ ¼ KP . Derive the transfer function for the liquid-level control
RðsÞ
system.
Solution
The governing equations of the components of the control system are as follows.

dh
ðiÞ Tank: A ¼ qin  qout (a)
dt
g
ðiiÞ Valve: qout ¼ h (b)
R
ðiiiÞ Sensor: ys ¼ μ s h (c)
ðivÞ Error: e ¼ r  ys (d)
ðvÞ Controller: u ¼ KP e (e)
ðviÞ Actuator: qin ¼ μa u (f)

In Eq. (b), g is the gravitational acceleration. Taking Laplace transforms of Eqs. (a) to (f) and
performing substitutions, yields
 g  
As þ HðsÞ ¼ Qin ðsÞ ¼ μa KP RðsÞ  μs HðsÞ (g)
R

It follows that the transfer function of the control system is

HðsÞ μ a KP β
¼ ¼ (h)
RðsÞ As þ g þ μ μ KP Ts þ 1
s a
R
with
A μ KP
T¼g ; β¼g a (i)
þ μ s μ a KP þ μs μa KP
R R

CHAPTER SUMMARY
This chapter presents the basic concepts and working principles of feedback control systems.
A typical feedback control system (also known as closed-loop control system) consists of
four major components: the plant, sensor, controller, and actuator. A transfer function
formulation for feedback control systems with input, disturbance, and sensor noise is
654 8 Introduction to Feedback Control Systems

derived. Three advantages of closed-loop control, compared to open-loop control, are

demonstrated in examples. The PID control algorithm, which is widely used in applications,
and PID gain-tuning methods are introduced. In the transfer function formulation, transient
and steady-state behaviors, stability, and steady-state errors of closed-loop control systems
are analyzed. The root locus method, which is a commonly used tool for analysis and design
of closed-loop control system, is introduced. Finally, three examples of feedback control
systems in applications are presented.
Upon completion of this chapter, you should be able to:

(1) Understand the advantages of closed-loop (feedback) control, compared to open-loop

control.
(2) Obtain a transfer function formulation for a feedback control system.
(3) Analyze the performance of a feedback control system in terms of transient response,
stability, and steady-state error.
(4) Implement the PID control algorithm for feedback control.
(5) Use the root locus method to design feedback control systems.
(6) Compute the response of feedback control systems by MATLAB and Simulink.

REFERENCES
1. R. C. Dorf and R. H. Bishop, Modern Control Systems, 13th ed., Pearson, 2016.
2. G. Franklin, J. Powell, and A. Emami-Naeini, Feedback Control of Dynamic Systems, 8th ed.,
Pearson, 2018.
3. N. S. Nise, Control Systems Engineering, 8th ed., Wiley, 2020.

PROBLEMS
Section 8.1 General Concepts
8.1 Draw a block diagram for an open-loop system and a block diagram for a closed-loop
system, which are in applications of mechanical systems. Follow the formats in Figure
8.1.1.
8.2 Draw a block diagram for an open-loop system and a block diagram for a closed-loop
system, which are used in applications of electromechanical systems. Follow the formats
in Figure 8.1.1.
8.3 For the control system in Figure P8.3, derive the equations of the output Y(s), error E(s),
and actuator output V(s), in terms of the inputs R(s) and D(s). Also, obtain the relevant
transfer functions in each of the three cases, which are stated as follows:
 Problems 655

YðsÞ ¼ TR ðsÞRðsÞ þ TD ðsÞDðsÞ

EðsÞ ¼ TRE ðsÞRðsÞ þ TDE ðsÞDðsÞ
VðsÞ ¼ TRV ðsÞRðsÞ þ TDV ðsÞDðsÞ

D(s)
Controller Actuator Plant
R(s) + E(s) V(s) + + Y(s)
Gc (s) Ga (s) Gp (s)
−

Figure P8.3

Section 8.2 Advantages of Closed-Loop Control Systems

8.4 This problem concerns reduction of the effects of parameter variations by closed-loop
control. An open-loop system is shown in Figure P8.4(a), where the parameter k has a
variation Δk. Let the parameter value be k = 25. The required steady-state response of
the system to a unit step input is yss ¼ 1.
(a) In the ideal case of Δk ¼ 0 (no parameter variation), show that the open-loop system
meets the requirement of yss ¼ 1.
(b) For Δk ¼ 5%, 5% and 10% of k, plot the step responses of the open-loop system
in one figure, and determine whether the requirement of yss ¼ 1 is met in each
case.
(c) Consider the closed-loop system in Figure P8.4(b), where P control is implemented.
Show that, for a large control gain KP , the steady-state response yss of the control
system approaches 1 for Δk ≠ 0. This implies that P control can reduce the effects of
parameter variations.
(d) Let Δk be an unknown constant that is bounded by jΔk=kj < 1. A closed-loop
system with a PI controller is shown in Figure P8.4(c). Design the controller
parameters KP and KI to render yss ¼ 1. Make sure that the closed-loop system is
stable.
(e) Based on the results obtained in (d), choose appropriate values of KP and KI ,
and plot the step responses of the closed-loop system for Δk ¼ 5%, 5%
and 10% of k in one figure. By examining the system responses, determine
whether the effect of the parameter variation Δk can be eliminated by the PI
controller.
656 8 Introduction to Feedback Control Systems

R(s) 25 Y(s)
s2 + 4s + k + Δk

(a)

R(s) + 25 Y(s)
Kp
− s 2 + 4s + k + Δk

(b)

R(s) 1 25 Y(s)
+
Kp + KI
− s s2 + 4s + k + Δk

(c)

Figure P8.4

8.5 This problem concerns disturbance rejection by closed-loop control. An open-loop

system is shown in Figure P8.5(a), where a second-order system is subject to a
reference input R(s) and a disturbance D(s). Assume a constant distance, dðtÞ ¼ d0 .
The required steady-state response of the system subject to a unit step input ðrðtÞ ¼ 1Þ is
yss ¼ 1.
(a) In the ideal case of d0 ¼ 0 (zero disturbance), show that the open-loop system meets
the requirement of yss ¼ 1.
(b) For d0 ¼ 1; 2 and 5, plot the step responses of the open-loop system in one figure,
and determine whether the requirement of yss ¼ 1 is met in each case.
(c) With d0 ≠ 0, consider the closed-loop system in Figure P8.5(b), where a PI controller
is implemented. Show that the parameters KP and KI can be chosen to render the
desired steady-state response, yss ¼ 1. Make sure that the closed-loop system is
stable.
(d) Based on the results obtained in part (c), choose the values of KP and KI , and plot the
step responses of the closed-loop system d0 ¼ 1; 2 and 5, in one figure. By examining
the response curves, show that yss ¼ 1 in all the cases, and that the effect of the
disturbance is rejected by the closed-loop control.
 Problems 657

D(s)

R(s) + Y(s)
+ 25
s 2 + 4s + 25

(a) Open loop

D(s)

R(s) 1 + 25 Y(s)
+
Kp + KI
− s s 2 + 4s + 25

(b) Closed loop

Figure P8.5

Section 8.3 PID Control Algorithm

8.6 This problem concerns improvement of system response by closed-loop control. Figure
P8.6(a) shows a second-order open-loop system. It is desired to have a step response
(with r(t) = 1) that meets the following requirements: (i) the maximum overshoot Mp is
less than 30%; and (ii) the settling time ts (with a 2% criterion) is 2 s.
(a) Compute the maximum overshoot and settling time of the open-loop system.
(b) If the open-loop system does not meet the above-mentioned requirements, consider
the closed-loop system in Figure P8.6(b), where a PD controller is implemented.
Show that the controller parameters KP and KD can be chosen to satisfy the step-
response requirements.
(c) Plot the step response of the closed-loop system with the controller parameters
determined in part (b).

R(s) 25 Y(s)

s2 + 2s + 25

(a) Open loop

R(s) 25 Y(s)
+
2
− s + 2s + 25

Kp + KDs

(b) Closed loop

Figure P8.6
658 8 Introduction to Feedback Control Systems

8.7 Consider the unity-feedback system in Figure P8.7, where the plant is unstable, and a PD
controller is installed.
(a) Determine the conditions of the controller gain constants ðKP and KD Þ such that the
closed-loop system is stable.
(b) With the controller gain constants satisfying the stability conditions given in part (a),
determine whether the steady-state error ess of the control system exists for a step
input rðtÞ ¼ r0 . If so, compute the steady-state error.

Controller Plant
R(s) E(s) 4 Y(s)
+
Kp + KDs
− s(s − 1)(s + 2)

Figure P8.7

8.8 Consider the feedback control system in Figure 8.3.2, where a first-order system with
parameter T = 0.2 and b = 1 is under PID feedback control. Let the reference input and
disturbance be r(t) = 1 and d(t) = 0. Select the control gain constants KP ; KI ; and KD
such that the step response of the closed-loop system meets the following control
objectives:
(a) the closed-loop system is stable
(b) the settling time of the output (5% criterion) is less than 1.5 seconds
(c) the maximum overshoot of the output is less than 15%
Also, with the determined control gain constants, plot the response y(t) of the control
system.
8.9 Consider a mass–damper system described by the transfer function

YðsÞ 1
GðsÞ ¼ ¼
FðsÞ sðs þ 4Þ

Let the mechanical system be the plant, with force f as the input and the displacement y
as the output.
(a) Build a Simulink model of the control system shown in Figure P8.9, where the plant
is controlled by a PID controller. Here, the reference input is r(t) = 1, and the control
gain constants initially are set as KP ¼ 10, KI ¼ 0, and KD ¼ 0.
(b) With the model obtained in part (a) and by Table 8.3.1, tune the values of the control
gain constants to meet the following control objectives:
 Problems 659

(i) the closed-loop system is stable

(ii) the displacement y reaches the desired value (y = 1) in less than 2 s
(iii) the maximum overshoot of the response y is less than 10%
With the control gain constants selected, plot the response of the control system.

Kp +
Input Output
1 f(t)
r(t) +− KI + Plant y(t)
s
du Subsystem
KD +
dt

Figure P8.9

8.10 Consider the armature-controlled DC motor in Example 6.5.4, with its parameters
given in Example 6.6.2. Let the motor be the plant (dynamic system to be controlled),
with the applied voltage vin as the input and the motor speed ω ¼ θ_ as the output. Here,
the load TL is assumed to be zero.
(a) Build a Simulink model of the motor speed control system with a PID controller, as
shown in Figure P8.9, where f ¼ vin , y ¼ ω, and the reference input is
rðtÞ ¼ 40π rad=s (or 1200 rpm), which is a step function. For the model, the control
gain constants are initially set as KP ¼ 2, KI ¼ 0, and KD ¼ 0.
(b) With the model obtained in (a) and by Table 8.3.1, tune the values of the control
gain constants to meet the following control objectives:
(i) the closed-loop system is stable
(ii) the motor reaches the desired speed ð40π rad=sÞ in less than 3 s
(iii) the maximum overshoot of the motor speed is less than 25%
With the control gain constants selected, plot the response of the control system.
8.11 Consider a unity-feedback system in Figure 8.3.4, where the plant transfer function is

4
Gp ðsÞ ¼ (a)
sðs2 þ 8s þ 25Þ

and a PD controller is implemented. Determine the control gains KP and KD by the

Ziegler–Nichols ultimate-cycle method and plot the response of the control system
subject to a unit step input rðtÞ ¼ 1.
660 8 Introduction to Feedback Control Systems

8.12 Consider a unity-feedback system in Figure 8.3.4, where the plant transfer function is

3ðs þ 1Þ
Gp ðsÞ ¼ (a)
ðs2 þ 4s þ 5Þðs2 þ 8s þ 25Þ

and a PID controller is implemented. Determine the control gains KP , KI , and KD by

following the steps of the Ziegler–Nichols ultimate-cycle method in Section 8.3.2. Also,
with the determined control gains, plot the response of the control system subject to a
unit step input rðtÞ ¼ 1.

Section 8.4 Control System Analysis

8.13 Consider the closed-loop system in Figure 8.1.3(b), with

1 10
Gp ðsÞ ¼ ; Gs ðsÞ ¼ 1; Gc ðsÞ ¼ 10 þ 2s; Ga ðsÞ ¼
2s þ 1 0:2s þ 1

(a) Derive the transfer functions TR ðsÞ; TD ðsÞ; and TN ðsÞ in Eq. (8.1.2) and determine
the poles and zeros of the transfer functions.
(b) Plot the response yðtÞ of the system with inputs rðtÞ ¼ 2δðtÞ, d = 1. and n = 0, where
δðtÞ is the Dirac delta function.
8.14 Consider the closed-loop system in Figure 8.1.3(b), with

1 2
Gp ðsÞ ¼ ; Gs ðsÞ ¼ 1; Gc ðsÞ ¼ 5 þ ; Ga ðsÞ ¼ 8
s2 þ 4s þ 16 s

(a) Derive the loop transfer function L(s) as defined in Eq. (8.1.5), obtain the poles of
L(s), and determine the closed-loop poles.
(b) Find the steady-state response yss of the system subject to rðtÞ ¼ 4 and d = n = 0.
8.15 Consider the unity-feedback control system in Figure P8.15, where G(s) is the open-
loop transfer function. For each of the following open-loop transfer functions, derive
the stability conditions of the corresponding closed-loop system in terms of the
parameters.

s4
System 1: GðsÞ ¼ K
s2  4s þ 25
as þ b
System 2: GðsÞ ¼ 3
s þ 3s2  5s  7

1 1
System 3: GðsÞ ¼ KP þ KI þ KD s 2
s s þ 16
 Problems 661

R(s) + E(s) Y(s)

G(s)
−

Figure P8.15

8.16 Consider the unity-feedback control system in Figure P8.15, where G(s) is the open-
loop transfer function. For each of the following open loop transfer functions, find the
steady-state error ess of the corresponding closed-loop system, if any. Note that steady-
state error may not always exist.

3
System 1: GðsÞ ¼ , rðtÞ ¼ 5ð1  e2t Þ
s þ 10
sþ1
System 2: GðsÞ ¼ K ; with K > 10, rðtÞ ¼ 5
s5
sþ2
System 3: GðsÞ ¼ 3 2 , rðtÞ ¼ 2t
s ðs þ 3Þ

1 1
System 4: GðsÞ ¼ 3 þ 2 þ s 2 , rðtÞ ¼ 5t
s s þ4
8.17 In Figure P8.17, a second-order plant is under P control with gain K.
(a) Determine the range of K for stability of the closed-loop system.
(b) Determine the value of K such that the step response of the closed-loop system has a
maximum overshoot Mp ¼ 10%.
(c) With the K value in part (b), determine the steady-state error of the closed-loop
system subject to a ramp input rðtÞ ¼ 2t.
(d) With the K value determined in part (b), plot the step response of the closed-loop
system by MATLAB.

Plant
Controller
R(s) + 3 Y(s)
K
− s(s + 25)

Figure P8.17
662 8 Introduction to Feedback Control Systems

8.18 A dynamic system is governed by the coupled differential equations

y€ þ 2y_ þ ðλ  3Þy þ μz ¼ f
z_ þ z  y ¼ 0

where f is the input, z is the output, and λ and μ are constant parameters to be designed.

(a) Obtain the transfer function GðsÞ ¼ ZðsÞ=FðsÞ.

(b) With the GðsÞ of part (a), derive the conditions on λ and μ for the system to be
stable.

Section 8.5 The Root Locus Method

8.19 A unity-feedback control system, as shown in Figure 8.3.4, has a loop transfer function

Kðs þ 3Þ
LðsÞ ¼ Gc ðsÞGp ðsÞ ¼
ðs  1Þðs þ 2Þðs þ 4Þ

Sketch the root locus for K > 0 by following the six steps in Section 8.5.3.
8.20 A unity-feedback control system, as shown in Figure 8.3.4, has a loop transfer function

K
LðsÞ ¼ Gc ðsÞGp ðsÞ ¼
sðs2 þ 8s þ 25Þ

Sketch the root locus for K > 0 by following the six steps in Section 8.5.3.
8.21 A feedback control system has the following characteristic equation:

sþ1
1þ ¼0
ðs þ 3Þðs2 þ Ks þ 4Þ

Sketch the root locus for K > 0 by following the six steps in Section 8.5.3.
8.22 Consider the unity-feedback control system in Problem 8.20.
(a) Sketch the root locus for K > 0 by MATLAB.
(b) Determine the range of K for the stability of the control system.
(c) Determine the gain K such that the complex closed-loop poles have a damping
ratio of 0.5.
(d) With the gain of part (c), find the actual maximum overshoot and peak time by
plotting the system step response.
8.23 Consider the unity-feedback control system in Problem 8.21.
(a) Sketch the root locus for K > 0 by MATLAB.
(b) Determine the range of K for the stability of the control system.
(c) Determine the gain K such that the complex closed-loop poles have a damping
ratio of 0.707.
(d) Determine the break-in point of the root locus and the gain.
 Problems 663

8.24 A unity-feedback control system, as shown in Figure 8.3.4, has a loop transfer function

Kðs þ 2Þ
LðsÞ ¼ Gc ðsÞGp ðsÞ ¼
sðs þ 8Þðs  2Þ

(a) Sketch the root locus for K > 0 by MATLAB.

(b) Determine the range of K for the stability of the control system.
(c) Determine the gain K such that the complex closed-loop poles have a damping ratio
of 0.2.
(d) With the gain of part (c), find the actual maximum overshoot and settling time (with
a 2% criterion) by plotting the system step response.

Section 8.6 Analysis and Design by the Root Locus Method

8.25 A closed-loop control system has the following characteristic equation:

Kðs þ 2Þ
1þ ¼0
ðs þ 3Þðs  1Þðs2 þ 6s þ 13Þ

(a) Determine the range of gain K for stability by using the Routh–Hurwitz stability
criterion.
(b) Verify the stability result in part (a) by sketching the root locus by MATLAB.
8.26 Repeat Problem 8.12 using the root locus method.
8.27 Consider the feedback control system in Figure P8.27, where a third-order plant is
under PD control. With the dominant complex poles, determine the gain K such that
the system response subject to an input rðtÞ ¼ 1 meets the following requirements: (i)
the settling time (with a 2% criterion) is less than 2 s, and the maximum overshoot is not
larger than 20%. Also, with the determined K, plot the step response of the system and
identify the actual settling time and maximum overshoot.

PD Controller Plant
R(s) + 1 Y(s)
K(s + 2)
− (s + 8)(s 2 + 4s + 8)

Figure P8.27
664 8 Introduction to Feedback Control Systems

8.28 Consider the feedback control system in Figure P8.28. Design a PID controller by zero
placement as shown in Section 8.6.3 such that, under a unit step input rðtÞ ¼ 1, the
response of the closed-loop system satisfies the following requirements:
(i) the steady-state error ess ¼ 0
(ii) the settling time (5% criterion) ts ≤ 5 s
(iii) the maximum overshoot Mp ≤ 10%

Controller Plant
R(s) + KI 1 Y(s)
Kp + + KDs
s s 2 + 2s + 2
−

Figure P8.28

Section 8.7 Additional Examples of Control Systems

8.29 A hoist control system is shown in Figure P8.29, where a pulley with an unextendible rope
carries a load (weight Mg); the pulley is connected to a shaft by a torsional spring (k); and
the shaft that is supported by a bearing (b) is driven by a torque τ applied by a motor for
lifting or lowering the load. The height y of the weight is measure by a sensor, with output
ys ¼ μs y, where μs is a gain constant. A controller determines a control signal u based on
the error e ¼ r  ys , where r is a reference input, representing the desired height of the
weight. The amplifier provides a motor voltage v by v ¼ μa u, where μa is a gain
dτ
constant. The motor torque τ is governed by the differential equation þ ατ ¼ βv,
dt
where α and β are constants. Let the mass moment of inertia of the motor rotor-shaft
assembly be J. Consider PID control.
(a) Write down the equations for all the components of the control system.
(b) Derive the transfer function of the control system from the reference input r to the
height y.
(c) Draw a block diagram of the control system, with r and Mg as the inputs and y and
θ_ 1 as the outputs

Pulley
Motor τ J θ1 × k
θ2
v
× g
Shaft
b
Io ,R
Amplifier

u y
ys Sensor
e − M
Controller
+

Figure P8.29
 Problems 665

8.30 A liquid-level control system is shown in Figure P8.30, where the height h of a tank of
constant area A is measured by a sensor; a controller determines a control signal u
based on error e ¼ r  ys ; and with u, a valve actuator adjusts the input volume flow
rate qin such that the tank height eventually reaches a desired level specified by the
reference input r. The sensor and actuator outputs are given by ys ¼ μs h and qin ¼ μa u,
respectively, with μs and μa being constants. Consider PI control.
(a) Write down the equations for all the components of the control system.
(b) Draw a block diagram for the control system, with r as the input and h and qout as
the outputs.
(c) Derive the transfer function HðsÞ=RðsÞ for the control system.
(d) Obtain a state-space model by using the equations obtained in part (a), with r as the
input, and h and qout as the outputs. Hint: with the PI controller, the control system
has two state variables, and you may choose the state variables as x1 ¼ h and
x2 ¼ u  KP e.

Valve
actuator
From reservoir

u pa
Controller qin

e
r ys Sensor

+ −

h pa

A R qout

Valve

Figure P8.30
9 Application Problems

Contents
9.1 Vibration Analysis of a Car Moving on a Bumpy Road 667
9.2 Speed Control of a Coupled Engine–Propeller System 686
9.3 Modeling and Analysis of a Thermomechanical System (a Bimetallic Strip
Thermometer) 697
9.4 Modeling and Analysis of an Electro-Thermo-Mechanical System
(a Resistive-Heating Element) 702
9.5 Feedback Control of a Liquid-Level System for Water Purification 711
9.6 Sensors, Electroacoustic, and Piezoelectric Devices 723
References 756

This chapter assembles six problems of combined dynamic systems from engineering appli-
cations, namely, vibration analysis of a moving car, speed control of a coupled engine–
propeller system (electromechanical system), modeling and analysis of a bimetallic strip
thermometer (thermomechanical system), modeling and analysis of a resistive-heating elem-
ent (electro-thermo-mechanical system), feedback control of a water purification system
component (liquid-level system), and certain working principles of sensors, electroacoustic,
and piezoelectric devices. Each problem is presented in one section.
The purpose of this ending chapter of the book is two-fold. First, the chapter serves as
a comprehensive review of the contents covered in the previous chapters, including the three-
key modeling technique (fundamental principles, basic elements, and ways of analysis); the
formulation of combined systems by transfer functions, block diagrams, and state-space
representations; the basic concepts of dynamic response analyses (time response, stability,
and frequency response); the necessity of feedback control systems; and the utility of the
software packages MATLAB/Simulink and Mathematica in simulation. Second, the chapter
gives the reader a taste of the model-based design of machines, devices, high-tech products,
and industrial processes. Indeed, the techniques of modeling, simulation, and analysis of
dynamic systems introduced in this text, with extension, are applicable to many practical
problems in engineering applications.

666
 9.1 Vibration Analysis of a Car Moving on a Bumpy Road 667

9.1 Vibration Analysis of a Car Moving on a Bumpy Road

In Chapter 3, a one-degree-of-freedom model of an automotive suspension system, which is
called a quarter-car suspension model, is derived (see Example 3.2.4). However, a quarter-
car model misses the rotation of the car body, which is important in vehicle dynamics. In this
section, a generalized model, which describes the rotation of the car body and the vibration
of the car engine, is presented. With this model, the dynamic response of a moving car subject
to two types of road condition is investigated.

9.1.1 System Description

Figure 9.1.1 shows a three-degrees-of-freedom (3-DOF) model that characterizes the motion
of a car moving on a road at a constant speed vc . This model is called a half-car model because
two tires of the car are considered. In the figure, the two pairs of spring and damper (k and c)
model the tires and suspension system; the rigid body ðMb ; Ib Þ represents the body of the car
including the chassis, with Ib being the moment of inertia of the body with respect to its center
of mass (point G); the lumped mass ðme Þ is the engine assembly that is mounted on the chassis
at point E by the mounts that are characterized by the spring and damper pair ðke ; ce Þ; l1 and
l2 are the distances of points A and B, where the tire-suspension assemblies are attached to

xe
g me
Engine
ke ce

Body xb
E
vc
Mb , Ib
θb
G
B A

Tire and k c k C Tire and

suspension l2 l1 suspension

y2 Road y1

Figure 9.1.1 A half-car model

668 9 Application Problems

the chassis, from the mass center G, respectively; and g is the gravitational acceleration. Here,
xb and θb are the translation and rotation of the car body; xe is the displacement of the engine
assembly; and y1 and y2 are the displacement excitations to the car, which are caused by the
interaction between the tires and the road surface.
For simplicity in modeling and analysis, we assume a small rotation angle θb of the car
body: jθb j << 1, which is true in normal car operating scenarios. Also, we assume that the
tires are in touch with the road surface all the time. To describe a temporary separation of the
tires and road surface, a complicated nonlinear dynamic model is necessary, which is beyond
the scope of this text.
The condition of a road surface can be described by the displacement excitations y1 and y2 .
For a horizontal and smooth road surface (ideal case), y1 ¼ y2 ¼ 0. For a road surface of
sinusoidal profile, the displacement excitations can be described by sinusoidal functions,
such as

y1 ðtÞ ¼ y0 sinðωtÞ
(9.1.1)
y2 ðtÞ ¼ y0 sinðωðt  t12 ÞÞ

where y0 is the amplitude of the surface profile, and

vc l 1 þ l2
ω¼ ; t12 ¼ (9.1.2)
Lr vc

with Lr being the characteristic length of the road and t12 the travel time for the distance
between the two tires (the distance between point A and point B). The parameter Lr describes
how bumpy the road surface is. For instance, for Lr ¼ 10 m, the car goes up and down on the
road surface for one cycle when traveling a distance of 2πLr ¼ 62:83 m. Therefore, the
smaller the Lr is, the bumpier the road is.
Potholes in a road, usually asphalt pavement, can cause severe damage to cars, including
tire, wheel, and bearing damage, misalignment, damage to the car body and suspension
systems, and harm to engine parts. A pothole of rectangular shape as shown in
Figure 9.1.2(a) can be described by the displacement excitations as follows

Figure 9.1.2 Road surface: (a) a pothole; (b) a bump

 9.1 Vibration Analysis of a Car Moving on a Bumpy Road 669

h  i
y1 ðtÞ ¼ hp uðt  t0 Þ  u t  ðt0 þ wp =vc Þ
h    i (9.1.3)
y2 ðtÞ ¼ hp u t  ðt0 þ t12 Þ  u t  ðt0 þ t12 þ wp =vc Þ

where hp and wp are the depth and width of the pothole, respectively; t0 is the time when the
front tire of the car (the right tire in Figure 9.1.1) hits the pothole; and t12 is given in Eq.
(9.1.2). Of course, a pothole of arbitrary shape can be similarly treated if the shape profile is
known.
A road with bumps can be also be modeled via the proper assignment of the displacement
excitations. For instance, a rectangular bump in road surface, as shown in Figure 9.1.2(b),
can be described by
h  i
y1 ðtÞ ¼ hb uðt  t0 Þ  u t  ðt0 þ wb =vc Þ
h    i (9.1.4)
y2 ðtÞ ¼ hb u t  ðt0 þ t12 Þ  u t  ðt0 þ t12 þ wb =vc Þ

where hb and wb are the height and width of the bump; and t0 is the time when the front tire of
the car hits the bump.

9.1.2 Dynamic Modeling

In the development of the half-car model shown in Figure 9.1.1, three independent displace-
ment parameters are selected to describe the motion of the 3-DOF system. Note that the
selection of independent displacements for a dynamic system is not unique. Listed below are
two sets of independent displacement parameters:

Set 1: xb , θb , and xe
Set 2: x1 , x2 , and xe

where the geometric meaning of the parameters in Set 1 has been explained previously; and x1
and x2 are the transverse displacements of the rigid body at points A and B (see Figure 9.1.1),
respectively, in the upward direction.
These two sets of parameters are equivalent in the sense that the parameters in one set can
be expressed by those in the other set. For instance, with the small rotation assumption
(jθb j << 1), Figure 9.1.3 indicates that

xb ¼ x1  l1 θb ¼ x2 þ l2 θb

It follows that the two sets of displacement parameters are related by

670 9 Application Problems

θb A
G

B
x2 xb x1

l2 l1

Figure 9.1.3 Rigid body of the half-car model with displacement parameters

x1 ¼ xb þ l1 θb ; x2 ¼ xb  l2 θb (9.1.5a)

or
l 2 x 1 þ l1 x 2 x1  x2
xb ¼ ; θb ¼ (9.1.5b)
l 1 þ l2 l 1 þ l2

For convenience of modeling and analysis, these two sets of parameters will be used
alternatively.
As mentioned in Sections 3.2 and 3.5, the equations of motion for a mechanical system can
be derived with two approaches: (i) the Newtonian approach, in which Newton’s laws are
applied to the free-body diagrams of the system; and (ii) the Lagrangian approach, in which
Lagrange’s equations are applied to the energy functions of the system. For the current
problem, both approaches are used.

Modeling by the Newtonian Approach

The free-body diagrams of the car are shown in Figure 9.1.4, where fsi and fdi , for i = 1, 2, 3,
are the internal forces of the springs and dampers, respectively. By Newton’s second law and
with the free-body diagrams, the governing equations of the rigid body and lumped mass are

For Mb Mb x€b ¼ Mb g  ð fs1 þ fd1 Þ  ð fs2 þ fd2 Þ þ ð fs3 þ fd3 Þ (9.1.6a)

For Ib Ib θ€b ¼ ð fs1 þ fd1 Þl1 þ ð fs2 þ fd2 Þl2 þ ð fs3 þ fd3 Þd (9.1.6b)
For me me x€e ¼ me g  ð fs3 þ fd3 Þ (9.1.6c)

where the moments of forces are with respect to the mass center G and in the positive
rotational direction. The internal forces of the system are given by
 9.1 Vibration Analysis of a Car Moving on a Bumpy Road 671

fs3 + fd3

d E
xb
xe
θb
A me
G

Mb ,Ib fs3 + fd3

Meg
fs1 + fd1
B Mbg

fs2 + fd2

Figure 9.1.4 Free-body diagrams of the half-car model

fs1 ¼ kðx1  y1 Þ; fd1 ¼ cðx_ 1  y_ 1 Þ

fs2 ¼ kðx2  y2 Þ; fd2 ¼ cðx_ 2  y_ 2 Þ (9.1.7)
fs3 ¼ ke ðxe  xE Þ; fd3 ¼ ce ðx_ e  x_ E Þ

where xE is the displacement of the rigid body at point E and it is given by

xE ¼ xb þ d  θ b (9.1.8)

With the parameters (xb , θb , xe ) of Set 1 and by using Eqs. (9.1.5), (9.1.7), and (9.1.8), Eq.
(9.1.6a) is written as

Mb x€b ¼  Mb g  kðx1  y1 Þ  cðx_ 1  y_ 1 Þ  kðx2  y2 Þ

 cðx_ 2  y_ 2 Þ þ ke ðxe  xE Þ þ ce ðx_ e  x_ E Þ
¼ Mb g  kðxb þ l1 θb  y1 Þ  cðx_ b þ l1 θ_ b  y_ 1 Þ  kðxb  l2 θb  y2 Þ
 cðx_ b  l2 θ_ b  y_ 2 Þ þ ke ðxe  xb  d  θb Þ þ ce ðx_ e  x_ b  d  θ_ b Þ

which, after rearrangement, becomes

Mb x€b þ ð2c þ ce Þx_ b  ce x_ e þ ðcl1  cl2 þ ce dÞθ_ b
(9.1.9a)
þ ð2k þ ke Þxb  ke xe þ ðkl1  kl2 þ ke dÞθb ¼ cðy_ 1 þ y_ 2 Þ þ kðy1 þ y2 Þ  Mb g

Similarly, Eqs. (9.1.6b) and (9.1.6c) are reduced to

Ib θ€b þ ðce d2 þ cl12 þ cl22 Þθ_ b þ ðce d þ cl1  cl2 Þx_ b  ce d x_ e

(9.1.9b)
þ ðke d2 þ kl12 þ kl22 Þθb þ ðke d þ kl1  kl2 Þxb  ke dxe ¼ cl1 y_ 1  cl2 y_ 2 þ kl1 y1  kl2 y2

me x€e þ ce ðx_ e  x_ b  d  θ_ b Þ þ ke ðxe  xb  d  θb Þ ¼ me g (9.1.9c)

Equations (9.1.9a) to (9.1.9c) are the governing equations of motion for the half-car model.
672 9 Application Problems

For convenience in analysis and simulation, Eqs. (9.1.9a) to (9.1.9c) are cast in vector-
matrix form as follows

x þ Cx_ þ Kx ¼ f
M€ (9.1.10)

where x and f are the displacement vector and the external force vector given by
0 1 0 1
xb cðy_ 1 þ y_ 2 Þ þ kðy1 þ y2 Þ  Mb g
x ¼ @θb A; f ¼ @ cl1 y_ 1  cl2 y_ 2 þ kl1 y1  kl2 y2 A (9.1.11)
xe me g

and M; C; and K are the mass, damping, and stiffness matrices, respectively, which are of the
form
2 3 2 3
Mb 0 0 ð2c þ ce Þ cðl1  l2 Þ þ ce d ce
M ¼ 4 0 Ib 0 5; C ¼ 4cðl1  l2 Þ þ ce d ce d2 þ cl12 þ cl22 ce d 5
0 0 me ce ce d ce x_ e
2 3 (9.1.12)
ð2k þ ke Þ kðl1  l2 Þ þ ke d ke
6 7
K ¼ 4kðl1  l2 Þ þ ke d ke d2 þ kl12 þ kl22 ke d 5
ke ke d ke

Note that M; C; and K are symmetric matrices.

Modeling by the Lagrangian Approach

The kinetic energy T, potential energy V, and Rayleigh dissipation function R of the car are
obtained as follows:

1 1 2 1
T ¼ Mb x_ 2b þ Ib θ_ b þ me x_ 2e
2 2 2
1 1 1
V ¼ kðxb þ l1 θb  y1 Þ2 þ kðxb  l2 θb  y2 Þ2 þ ke ðxb þ d  θb  xe Þ2
2 2 2 (9.1.13)
þ Mb g xb þ me g xe
1 1 1
R ¼ cðx_ b þ l1 θ_ b  y_ 1 Þ2 þ cðx_ b  l2 θ_ b  y_ 2 Þ2 þ ce ðx_ b þ d  θ_ b  x_ e Þ2
2 2 2

where Eqs. (9.1.5a) and (9.1.8) have been used.

With the energy functions given in Eq. (9.1.13), the governing equations of the car can be
obtained by Lagrange’s equations as given in Section 3.5. Note that the external nonconser-
vative forces are all zeros, namely, Qxb ¼ Qθb ¼ Qxe ¼ 0. This is because the gravitational,
spring, and damping forces have been included in the energy functions V and R. For xb , the
Lagrange’s equation is
 9.1 Vibration Analysis of a Car Moving on a Bumpy Road 673


d ∂T ∂T ∂R ∂V
 þ þ ¼0
dt ∂x_ b ∂xb ∂x_ b ∂xb

or
d
ðMb x_ b Þ þ cðx_ b þ l1 θ_ b  y_ 1 Þ þ cðx_ b  l2 θ_ b  y_ 2 Þ þ ce ðx_ b þ d  θ_ b  x_ e Þ
dt
þ kðxb þ l1 θb  y1 Þ þ kðxb  l2 θb  y2 Þ þ ke ðxb þ d  θb  xe Þ þ Mb g ¼ 0

which is the same as Eq. (9.1.9a). Also, for θb , the Lagrange’s equation is

d ∂T ∂T ∂R ∂V
 þ þ ¼0
_
dt ∂θ b ∂θb ∂θ_ b ∂θb

or
d _ 
Ib θ b þ cl1 ðx_ b þ l1 θ_ b  y_ 1 Þ  cl2 ðx_ b  l2 θ_ b  y_ 2 Þ þ d ce ðx_ b þ d  θ_ b  x_ e Þ
dt
þ kl1 ðxb þ l1 θb  y1 Þ  kl2 ðxb  l2 θb  y2 Þ þ d ke ðxb þ d  θb  xe Þ ¼ 0

which is the same as Eq. (9.1.9b). Finally, for xe , the Lagrange’s equation is

d ∂T ∂T ∂R ∂V
 þ þ ¼0
dt ∂x_ e ∂xe ∂x_ e ∂xe

or
d
ðme x_ e Þ  ce ðx_ b þ d  θ_ b  x_ e Þ  ke ðxb þ d  θb  xe Þ þ me g xe ¼ 0
dt

which is the same as Eq. (9.1.9c).

Thus, we have shown that the same equations of motion for the half-car model can be
obtained by both the Newtonian and Lagrangian approaches.

9.1.3 State-Space Representations

To determine the time response of the car, a state-space representation is established.
Because Eqs. (9.1.9a) to (9.1.9c) are a set of three second-order differential equations, six
state variables are selected as follows:
0 1
 z1
x B .. C
z¼ ¼@. A (9.1.14)
x_
z6

where x is the displacement vector given in Eq. (9.1.11).

674 9 Application Problems

According to Eq. (9.1.10),

d
_ ¼€
ðxÞ x ¼ M1 ðf  Cx_  KxÞ
dt
d
Combining the previous equation with ðxÞ ¼ x_ results in the following state equation in
dt
matrix form:

z_ ¼ Az þ Bf (9.1.15)

with
0 I 0
A¼ ; B¼ (9.1.16)
M1 K M1 C M1

The assignment of the outputs in a state-space representation depends on the

particular interest and requirement in the dynamic analysis. In the current problem,
we are interested in the transverse vibration of the car and the resultant internal
forces of the suspension system and engine mounts. Thus, six outputs are specified as
follows

y1 ¼ xb ; y2 ¼ θb ; y3 ¼ xe ;
y4 ¼ fs1 þ fd1 ¼ kðxb þ l1 θb  y1 Þ þ cðx_ b þ l1 θ_ b  y_ 1 Þ
(9.1.17)
y5 ¼ fs2 þ fd2 ¼ kðxb  l2 θb  y2 Þ þ cðx_ b  l2 θ_ b  y_ 2 Þ
y6 ¼ fs3 þ fd3 ¼ ke ðxe  xb  d  θb Þ þ ce ðx_ e  x_ b  d  θ_ b Þ

where Eqs. (9.1.5a), (9.1.7), and (9.1.8) have been used. It follows from Eqs. (9.1.14) and
(9.1.17) that the output equations of the car are as follows:

y1 ¼ z1 ; y2 ¼ z2 ; y3 ¼ z3
y4 ¼ kðz1 þ l1 z2 Þ þ cðz4 þ l1 z5 Þ  cy_ 1  ky1
(9.1.18)
y5 ¼ kðz1  l2 z2 Þ þ cðz4  l2 z5 Þ  cy_ 2  ky2
y6 ¼ ke ðz3  z1  d  z2 Þ þ ce ðz6  z4  d z5 Þ

Equation (9.1.18) is cast in matrix form as follows

y ¼ Dz þ p (9.1.19)

where y is the output vector given by

y ¼ ð y1 y2 . . . y6 Þ T (9.1.20a)
 9.1 Vibration Analysis of a Car Moving on a Bumpy Road 675

and
2 3 0 1
1 0 0 0 0 0 0
6 0 1 0 0 0 07 B 0 C
6 7 B C
6 0 0 1 0 0 07 B 0 C
D¼6
6 k
7; p ¼ B C (9.1.20b)
6 kl1 0 c cl1 07 7
Bcy_ 1  ky1 C
B C
4 k kl2 0 c cl2 05 @cy_  ky2 A
2
ke ke d ke ce ce d ce 0

Thus, a state-space representation for the half-car model in Figure 9.1.1, which consists of
the state equation (9.1.15) and the output equation (9.1.19), has been established. The
solution of these equations gives the dynamic response of the car traveling on a road.

9.1.4 Quasi-Static Deflections and Dynamic Response

For convenience of analysis and simulation, the effect of gravity and that of road conditions are
separated. To this end, consider the following quasi-static situation: the car experiences no road
excitations (y1 ¼ y2 ¼ 0), and it has no vibration (x_ ¼ 0). Under these circumstances, the
deflections of the car, according to Eq. (9.1.10), satisfy the following equilibrium equation:

Kxqs ¼ f g (9.1.21)

with
0 1 0 1
ex b Mb g
xqs ¼ @eθ b A; f g ¼ @ 0 A (9.1.22)
exe me g

Here, xqs contains the quasi-static deflections of the car under the gravitational forces, and it
is given by

xqs ¼ K1 f g (9.1.23)

Note that the quasi-static deflections are independent of the car speed.
Write

x ¼ xqs þ u; f ¼ f g þ q (9.1.24)

where u contains the displacements measured from the quasi-static equilibrium configur-
ation, and q only contains the displacement excitations. Equation (9.1.24) means that the
total response of the car is the sum of its quasi-static response by gravity and dynamic
response by road excitation. Substituting Eq. (9.1.24) into Eq. (9.1.10), and using Eq.
(9.1.21), yields the governing equation of the car dynamic response as follows:

Mu€ þ Cu_ þ Ku ¼ q (9.1.25)

676 9 Application Problems

The resultant internal forces of the car suspension systems can be obtained by Eq. (9.1.17),
with the solution of Eq. (9.1.25) and the expression in Eq. (9.1.24).
In the subsequent discussion, the symbol u for the dynamic response is replaced by x for
simplicity.

9.1.5 Steady-State Vibration of a Car Moving on Road of Sinusoidal Profile

In this section, the dynamic response of the car moving on a road with the sinusoidal surface
profile described by Eq. (9.1.1) is determined. In particular, the steady-state vibration of the
car excited by such displacement excitations is studied. As mentioned in Section 9.1.4, the
effect of gravity is neglected for dynamic analysis. The quasi-static deflections can always be
added by using Eq. (9.1.23).
In formulating the vibration problem, substitute Eq. (9.1.1) into Eq. (9.1.25), to obtain
n  
x þ Cx_ þ Kx ¼ y0 kq1 sinðωtÞ þ kq2 sin ωðt  t12 Þ
M€
 
þ cωq1 cosðωtÞ þ cωq2 cos ωðt  t12 Þ g (9.1.26)
0 1 0 1
1 1
q1 ¼ @l1 A; q2 ¼ @l2 A (9.1.27)
0 0

where u has been replaced by x. Equation (9.1.26) can be written as

x þ Cx_ þ Kx ¼ Im y0 Q0 ðωÞejωt g
M€ (9.1.28)

with
 
Q0 ðωÞ ¼ ðk þ jcωÞ q1 þ ejωt12 q2 (9.1.29)
 
where cosθ ¼ Im jejθ has been used. Note that the frequency-dependent vector Q0 ðωÞ is
not a function of time. Thus, from the theory of complex numbers, the steady-state solution
of Eq. (9.1.26) is expressed by
 
xss ðtÞ ¼ Im yss ðtÞ (9.1.30)

where yss ðtÞ is the steady-state solution of the differential equation

y þ Cy_ þ Ky ¼ y0 Q0 ðωÞejωt
M€ (9.1.31)

The steady-state solution of Eq. (9.1.31), in the theory of differential equations (Section 2.6),
is of the form
 9.1 Vibration Analysis of a Car Moving on a Bumpy Road 677

yss ðtÞ ¼ Ye jωt (9.1.32)

Plugging Eq. (9.1.32) into Eq. (9.1.31) gives

Y ¼ y0 GðjωÞQ0 ðωÞ (9.1.33)

where the matrix GðjωÞ is a transfer function of the car, and it is given by
 1
GðjωÞ ¼ ω2 M þ jωC þ K (9.1.34)

It follows that the steady-state response of the car is given by

 
xss ðtÞ ¼ y0 Im GðjωÞQ0 ðωÞe jωt (9.1.35)

We write
0 1 0 1
xb;ss ðtÞ Xb ðωÞ
xss ðtÞ ¼ @θb;ss ðtÞ A; GðjωÞQ0 ðωÞ ¼ @Θb ðωÞ A (9.1.36)
xe;ss ðtÞ Xe ðωÞ

where xb;ss ðtÞ, θb;ss ðtÞ, and xe;ss ðtÞ are the steady-state displacements of the car; and Xb ðωÞ,
Θb ðωÞ, and Xe ðωÞ are complex functions of the excitation frequency ω. Thus, the steady-state
displacements of the car can be written as
 
xb;ss ðtÞ ¼ y0 jXb ðωÞj sin ωt þ ∠ Xb ðωÞ
 
θb;ss ðtÞ ¼ y0 jΘb ðωÞj sin ωt þ ∠Θb ðωÞ (9.1.37)
 
xe;ss ðtÞ ¼ y0 jXe ðωÞj sin ωt þ ∠ Xe ðωÞ

where jzj and ∠ z are the magnitude and phase angle of the complex function z, respectively.
According to Section 7.3, Xb ðωÞ, Θb ðωÞ, and Xe ðωÞ are the frequency response functions of the
car.
For numerical simulation, the values of the car parameters are chosen as follows:

Tire-suspension assemblies :
k ¼ 2:54  105 N=m; c ¼ 2:73  103 kg=s; l1 ¼ 1:46 m; l2 ¼ 1:34 m
Car body: Mb ¼ 1; 750 kg; Ib ¼ 2; 530 kg-m2
Engine and engine mounts:
me ¼ 250 kg; ke ¼ 1:32  105 N=m; ce ¼ 1:03  103 kg=s; d ¼ 1:65 m
678 9 Application Problems

Also, for the road condition, assume that y0 ¼ 0:05 m and Lr ¼ 3 m, which describe a bumpy
road.
With Eq. (9.1.37), the steady-state displacements of the car are plotted versus the car speed
in Figures 9.1.5 to 9.1.7, for 0 ≤ vc ≤ 50 m=s. In Figure 9.1.5, the amplitude and phase of
xb;ss ðtÞ are given by y0 jXb ðωÞj and ∠ Xb ðωÞ. The amplitudes and phases in Figure 9.1.6 and
9.1.7 are similarly obtained from Eq. (9.1.37). As observed from the figures, an increase in the
car speed leads to increased steady-state vibration amplitudes of the car.
Let the acceptable amplitudes of the car displacements, for operation safety and driver
comfort, be specified as follows:

jxe;ss ðtÞj ≤ 0:15 m; jθb;ss ðtÞj ≤ 6 deg; jxe;ss ðtÞj ≤ 0:15 m

As indicated by the dashed lines in Figures 9.1.5 to 9.1.7, the car speed must be under
22.25 m/s (80.1 km/h), in order to meet the requirement of car displacement amplitudes.
For a less bumpy road, say, y0 ¼ 0:05 m and Lr ¼ 5 m., the amplitudes of the steady-state
displacements of the car versus the car speed are plotted in Figure 9.1.8. It can be seen that
the upper limit of the car speed is 33.85 m/s (121.9 km/h) in order to meet the amplitude
requirement.

Steady-State Displacement of Car Body, xb

0.3
Amplitude (m)

0.2
0.15 m

0.1

0
0 5 10 15 20 25 30 35 40 45 50

−50
Phase (deg)

−100

−150

−200
0 5 10 15 20 25 30 35 40 45 50
Speed vc (m/s)

Figure 9.1.5 Steady-state displacement of the car body, for Lr ¼ 3 m

 9.1 Vibration Analysis of a Car Moving on a Bumpy Road 679

Steady-State Rotation of Car Body, θb

10

Amplitude (deg)
6 deg

0
0 5 10 15 20 25 30 35 40 45 50

0
Phase (deg)

−50

−100

−150
0 5 10 15 20 25 30 35 40 45 50
Speed vc (m/s)

Figure 9.1.6 Steady-state rotation of the car body, for Lr ¼ 3 m

Steady-State Displacement of Car Engine, xe

1
Amplitude (m)

0.5

0.15 m

0
0 5 10 15 20 25 30 35 40 45 50

−50
Phase (deg)

−100

−150

−200
0 5 10 15 20 25 30 35 40 45 50
Speed vc (m/s)

Figure 9.1.7 Steady-state displacement of the car engine, for Lr ¼ 3 m

680 9 Application Problems

Steady-State Displacements of Car

0.1

⏐xb,ss⏐(m)
0.08

0.06

0.04
0 5 10 15 20 25 30 35 40 45 50

3.5
⏐θb,ss⏐(deg)

2.5

2
0 5 10 15 20 25 30 35 40 45 50

0.25
⏐xe ,ss⏐(m)

0.2
0.15 m
0.15

0.1
0 5 10 15 20 25 30 35 40 45 50
Speed vc (m/s)

Figure 9.1.8 Steady-state displacement amplitudes of the car, for Lr ¼ 5 m

In generating the steady-state displacement plots shown in Figures 9.1.5 to 9.1.8, the
following MATLAB commands were used:

% Car parameters
me = 250; ke = 1.32e5; ce = 1.03e3; d = 1.65;
Mb = 1750; Ib = 2530;
k = 2.54e5; c = 2.73e3; l1 = 1.34; l2 = 1.46;

% Road condition
y0 = 0.05; Lr = 3;

% Parameter matrices
MM = diag([Mb Ib me]);
CC = [2*c+ce, ce*d+c*(l1-l2), -ce;
ce*d+c*(l1-l2), ce*d^2+c*(l1^2+l2^2), -ce*d;
-ce, -ce*d, ce];
KK = [2*k+ke, ke*d+k*(l1-l2), -ke;
ke*d+k*(l1-l2), ke*d^2+k*(l1^2+l2^2), -ke*d;
-ke, -ke*d, ke];
 9.1 Vibration Analysis of a Car Moving on a Bumpy Road 681

% Initialization
npts = 1001;
vc = linspace(0,50,npts); % Car speed range
omega = vc/Lr; % Excitation frequency
t12 = (l1+l2) ./ vc;
Amplitude_xe = zeros(npts,1);
Phase_Angle_xe = zeros(npts,1);
Amplitude_xb = zeros(npts,1);
Phase_Angle_xb = zeros(npts,1);
Amplitude_Thetab = zeros(npts,1);
Phase_Angle_Thetab = zeros(npts,1);
% Computation of dynamic response
q1 = [1; l1; 0];
q2 = [1; l2;0];
for index = 1:1:npts
omg = omega(index);
t0 = t12(index);
Q0 = (k+c*1 j*omg)*(q1+exp(-1 j*omg*t0)*q2);
xss = y0*inv(-omg^2*MM + 1 j*omg*CC + KK) * Q0;
Amplitude_xe(index,1) = abs(xss(3,1));
Phase_Angle_xe(index,1) = angle(xss(3,1));
Amplitude_xb(index,1) = abs(xss(1,1));
Phase_Angle_xb(index,1) = angle(xss(1,1));
Amplitude_Thetab(index,1) = abs(xss(2,1));
Phase_Angle_Thetab(index,1) = angle(xss(2,1));
end

9.1.6 Time Response of a Car Moving over a Pothole

In this section, the dynamic response of a car moving on a road surface containing a pothole
is investigated. As described by Eq. (9.1.3), for a pothole, vectors f and p in the state equation
(9.1.15) and output equation (9.1.19) become
0 1
0
0 1 B 0 C
cðv1 þ v2 Þ þ kðy1 þ y2 Þ B C
B 0 C
@ A B
f ¼ cl1 v1  cl2 v2 þ kl1 y1  kl2 y2 ; p ¼ B C (9.1.38)
B cv1  ky1 C C
0 @cv2  ky2 A
0
682 9 Application Problems

where
v1 ðtÞ ¼ hp ½δðt  t0 Þ  δðt  ðt0 þ wp =vc ÞÞ
(9.1.39)
v2 ðtÞ ¼ hp ½δðt  ðt0 þ t12 ÞÞ  δðt  ðt0 þ t12 þ wp =vc ÞÞ

with δðtÞ being the Dirac delta function. As discussed in Section 9.1.4, gravity is not included
in dynamic analysis because the quasi-static deflections of a car can be added by Eq. (9.1.23).
In addition, the initial condition
 T
zð0Þ ¼ z0 ¼ xb ð0Þ θb ð0Þ xe ð0Þ x_ b ð0Þ θ_ b ð0Þ x_ e ð0Þ (9.1.40)

needs to be specified in determination of the time response of the car.

Several analytical and numerical methods are available for solution of the state equation
(9.1.15). In this undergraduate text, the following two solution methods are introduced.

Method 1: Numerical Solution via Runge–Kutta Method

A fixed Runge–Kutta method of order four is presented in Section 2.8. Solution by this
method takes the following three steps.

(1) Assign a solution region in the time domain: 0 ≤ t ≤ tf , where tf is the final time of
simulation.
(2) Divide the solution region into N identical time intervals: tk ¼ kh, k = 0, 1, . . ., N, with the
time-step size h given by h ¼ tf =N. The step size h must small enough to assure the
convergence of the numerical solution.
(3) The state vector at tkþ1 , which is zkþ1 ¼ zðtkþ1 Þ, is given by the iterative formula

h
zkþ1 ¼ zk þ ð f 1 þ 2f 2 þ 2f 3 þ f 4 Þ; k ¼ 0; 1; 2; . . . N  1 (9.1.41)
6

where
f 1 ¼ Φðtk ; zk Þ

h h
f 2 ¼ Φ tk þ ; zk þ f 1
2 2
 (9.1.42)
h h
f 3 ¼ Φ tk þ ; zk þ f 2
2 2
f 4 ¼ Φðtk þ h; zk þ hf 3 Þ

with
Φðt; zðtÞÞ ¼ AzðtÞþBfðtÞ (9.1.43)

The MATLAB commands for implementing the above formula are given in Section 2.8.
 9.1 Vibration Analysis of a Car Moving on a Bumpy Road 683

When using the Dirac delta function, as shown in Eq. (9.1.39), the following approxima-
tion can be made
8
<1
; if tk ≤ t < tkþ1
δðtk  t Þ ¼ h (9.1.44)
:
0; otherwise

where t is a fixed nonnegative number.

Method 2: Solution by MATLAB and Mathematica

The state equation (9.1.15) can also be solved by using the Runge–Kutta solvers in software
packages, such as the MATLAB function ode45 and the Mathematica function NDSolve.
Refer to Section 2.8 for some demonstrative examples.
In numerical simulation, the same car parameters given in Section 9.1.5 are used. Assume
that the parameters of the pothole in Eq. (9.1.3) have the following values: hp ¼ 0:1 m and
wp ¼ 1 m. The car travels at speed vc ¼ 12 m=s (43.2 km/h or 26.8 mph), and hits the pothole
at time t = 4 s. Select the final time of simulation as tf ¼ 10 s. Let the initial conditions of the
car be:

xb ð0Þ ¼ 0:01 m; θb ð0Þ ¼ 0:005 rad; xe ð0Þ ¼ 0:02 m

x_ b ð0Þ ¼ 0; θ_ b ð0Þ ¼ 0; x_ e ð0Þ ¼ 0

The time response of the car is computed by Method 1 and the results on the car
displacements are plotted in Figure 9.1.9. As shown in the figure, the vibration of the car
caused by the initial disturbances is quickly dampened out. When the car hits the pothole at
t = 4 s, it experiences relatively large jumps in its displacements and rotation, especially the
displacement xe of the engine. After going over the pothole, it takes several seconds for the
vibration to settle down.
The resultant suspension forces are denoted by fRj ¼ fsj þ fdj , j = 1, 2, 3, where fsj and fdj are
the spring and damping forces given in Eq. (9.1.17). With the output equation (9.1.19), the
suspension forces are computed, and plotted against time in Figure 9.1.10. The upper subplot of
the figure shows a big jump in fR1 of the first suspension at t = 4 s when the front tire hits the
pothole, and shortly after, there is another jump in the opposite direction as the tire is about
to leave the pothole. A similar pattern of double jumps is found for fR2 of the rear tire, as
shown in the middle subplot. These two pairs of double jumps are separated by a time
interval t12 ¼ ðl1 þ l2 Þ=vc ¼ 0:233 s, which is clearly seen on comparing the upper and middle
subplots. For the internal force fR3 of the engine mounts, shown in the lower subplot, the
appearance of jumps is not as swift as for fR1 and fR2 . Note that the maximum amplitude of fR3
is 5.5 times as large as the weight of the engine (250 kgf = 2452.5 N), and that the impulses in
fR1 and fR2 are at least 25 times as large as the total weight of the car (2000 kgf = 19620 N). The
684 9 Application Problems

Transient Response of Car Moving over a Pothole

0.05

xb (m) 0

−0.05
0 1 2 3 4 5 6 7 8 9 10

2
θ b (deg)

−2
0 1 2 3 4 5 6 7 8 9 10

0.2

0.1
x e (m)

−0.1

−0.2
0 1 2 3 4 5 6 7 8 9 10

t (seconds)

Figure 9.1.9 Time response of the car moving over a pothole, at speed vc ¼ 12 m=s

high-amplitude displacements and impulsive internal forces shown in Figures 9.1.9 and
9.1.10 provide an explanation for car damage caused by potholes.
To generate the plots in Figures 9.1.9 and 9.1.10 by Method 1 (numerical solution via the
Runge–Kutta Method) a MATLAB function RKfun is created to obtain Φðt; zÞ in Eq.
(9.1.43), which is given as follows:
function z = RKfun(t, y, h_step, AA, BB, Par_Matrix)
% Purpose: to calculate A*y + B*f in dy/dt = A*y + B*f
% Here, t = time, y = state vector
% AA and BB = matrices A and B in Eq. (9.1.15)
% h_step = time step size in Eq. (9.1.37)
% Par_Matrix = [vc c k l1 l2 hb wb tb];
% Input the car parameters
 9.1 Vibration Analysis of a Car Moving on a Bumpy Road 685

Suspension Forces of Car Moving over a Pothole

×10 5
5

fR1 (N)

−5
0 1 2 3 4 5 6 7 8 9 10

×10 5
5
fR2 (N)

−5
0 1 2 3 4 5 6 7 8 9 10

×10 4
1

0.5
fR3 (N)

−0.5

−1

0 1 2 3 4 5 6 7 8 9 10
t (seconds)

Figure 9.1.10 Transient internal forces of the car moving over a pothole
vc = Par_Matrix(1); c = Par_Matrix(2); k = Par_Matrix(3);
l1 = Par_Matrix(4); l2 = Par_Matrix(5);
% Input the pothole parameters
hb = Par_Matrix(6); wb = Par_Matrix(7);
tb = Par_Matrix(8); % the time when the car hits the pothole
t12 = (l1+l2)/vc; % travel time between two suspensions

% Compute Heaviside and Dirac Delta functions for the pothole

tt1 = t – tb; tt1d = tt1 – wb/vc;
tt2 = tt1 – t12; tt2d = tt2 – wb/vc;
% To set heaviside(0) = 0. By default, MATLAB assigns heaviside(0) = 0.5.
if tt1 == 0, tt1 = tt1 – eps; end
686 9 Application Problems

if tt2 == 0, tt2 = tt2 – eps; end

if tt1d == 0, tt1d = tt1d – eps; end
if tt2d == 0, tt2d = tt2d – eps; end
% Heaviside functions
u1 = heaviside(tt1); u1d = heaviside(tt1d);
u2 = heaviside(tt2); u2d = heaviside(tt2d);
% Dirac delta functions
delta1 = 0; delta1d = 0; delta2 = 0; delta2d = 0;
if tt1 > 0 & tt1 <= h_step, delta1 = 1/h_step; end
if tt1d > 0 & tt1d <= h_step, delta1d = 1/h_step; end
if tt2 > 0 & tt2 <= h_step, delta2 = 1/h_step; end
if tt2d > 0 & tt2d <= h_step, delta2d = 1/h_step; end

% Compute f(t) given in Eq. (9.1.11)

f = zeros(3,1);
y1 = -hb*(u1 – u1d);
y2 = -hb*(u2 – u2d);
v1 = -hb*(delta1 – delta1d);
v2 = -hb*(delta2 – delta2d);
f(1) = c*(v1+v2) + k*(y1+y2);
f(2) = c*(l1*v1-l2*v2) +k*(l1*y1-l2*y2);

% Assign the function output

z = AA*y + BB*f;s

In the simulation, the function RKfun is used with MATLAB commands given in Section 2.8.

9.2 Speed Control of a Coupled Engine–Propeller System

In Example 3.3.1, the equations of motion of a rotor (motor–propeller assembly) are derived.
This type of dynamic system finds applications in helicopters and unmanned aerial vehicles.
In this section, feedback control of the rotation speed of the rotor is examined. To under-
stand the subsequent presentation, some knowledge on electromechanical systems
(Chapter 6) and feedback control systems (Chapter 8) is required.
The rotor system in consideration consists of a motor, a propeller, and an elastic shaft
connecting the motor and propeller. Figure 9.2.1(a) shows an open-loop system and
Figure 9.2.1(b) shows a closed-loop system for the rotor, where θp is the rotation angle of
the propeller, τm is the torque applied to the motor shaft by the electric motor, vm is a voltage
applied to the motor by the motor driver (amplifier), and u is the command input to the
driver. For the open-loop system, the rotation speed θ_ p of the propeller is controlled by
 9.2 Speed Control of a Coupled Engine–Propeller System 687

. .
Propeller θp Propeller θp

Sensor

ys

−
Elastic Elastic + r
shaft τm shaft τm

Electric Electric
motor motor
Controller

vm vm
u u
Driver Driver

(a) (b)

Figure 9.2.1 Feedback control of a rotor system (motor-propeller assembly): (a) open-loop system; and
(b) closed-loop system

a user-specified command input u. For the closed-loop system, the propeller rotation speed is
manipulated by u that is determined by a controller according to the error e between the
reference input r (desired rotation speed) and the sensor output ys (measured rotation speed).
In this study, a field-controlled direct-current (DC) motor is used for the rotor system.
The main purpose of implementing a feedback controller in the rotor system is to maintain
a desired rotation speed of the propeller under the influence of system parameter variations
and air damping. Below, open-loop and closed-loop systems are formulated, and the neces-
sity of feedback control is demonstrated through numerical simulations by using MATLAB
and Simulink.

9.2.1 Open-Loop System

The open-loop system in Figure 9.2.1(a) has four main parts: (a) a motor driver; (b) an
electrical subsystem; (c) a mechanical subsystem; and (d) a coupling mechanism between the
electrical and mechanical subsystems. A block diagram of the open-loop system with these
parts is shown in Figure 9.2.2.
The governing equations of the four parts of the open-loop system are as follows:

ðaÞ Motor driver: v m ¼ Ka u (9.2.1a)

where Ka is an amplifier gain.

688 9 Application Problems

Rotor (motor + propeller)

.
u vm Electrical if τm Mechanical θp
Ka Km
subsystem subsystem
Driver Coupling

Figure 9.2.2 Block diagram of the open-loop system

dif
ðbÞ Electrical subsystem: vm ¼ Rf if þ Lf (9.2.1b)
dt

where if is the current of the field circuit of the field controlled motor (see Section 6.4.2).

ðcÞ Mechanical subsystem: Jm θ€m þ bθ_ m þ kT θm  kT θp ¼ τm (9.2.1c)

Jp θ€p þ bD θ_ p sgnðθ_ p Þ þ kT θp  kT θm ¼ 0
2
(9.2.1d)

where θm is the rotation angle of the motor armature shaft; kT is the coefficient of a torsional
spring, modeling the elastic shaft connecting the motor to the propeller; and the term
bD θ_ p sgnðθ_ p Þ is a nonlinear drag torque due to air damping. Refer to Example 3.3.1 for
2

more details on the derivation of Eqs. (9.2.1c) and (9.2.1d).

Finally, there is

ðdÞ Electromechanical coupling : τ m ¼ Km if (9.2.1e)

where Km is the motor torque constant.

Ideal Case

In this case, it is assumed that the nonlinear damping torque bD θ_ p sgnðθ_ p Þ in Eq. (9.2.1d)
2

is negligible. This assumption, of course, is not applicable to those applications in

which aerodynamic forces play an essential role. However, solutions in the ideal case
can serve as a reference for dynamic analysis and control of the rotor system in
practical situations.
With this assumption, the open-loop system becomes linear with the transfer
function

Ωp ðsÞ k T Ka Km
Go ðsÞ ¼ ¼ (9.2.2)
UðsÞ ðRf þ Lf sÞ½Jm Jp s3 þ bJp s2 þ kT ðJm þ Jp Þs þ bkT 
 9.2 Speed Control of a Coupled Engine–Propeller System 689

where Ωp ðsÞ ¼ L½θ_ p  ¼ sΘp ðsÞ.

Denote the desired rotation speed of the propeller by Ω0 and consider a constant com-
mand input uðtÞ ¼ U0 . From Eq. (9.2.2), it can be shown that all the poles of sGo ðsÞUðsÞ have
negative real parts, satisfying the condition of the final-value theorem, Eq. (7.1.15). It follows
that the steady-state rotation speed is

Ka Km U0
ωp;ss ¼ lim θ_p ¼ lim sGo ðsÞUðsÞ ¼ (9.2.3)
t→∞ s→0 Rf b

By setting ðωp Þss ¼ Ω0 in Eq. (9.2.3), the command input for the system to reach the desired
speed is obtained as

Rf b
U0 ¼ Ω0 (9.2.4)
Ka Km

Equation (9.2.4) shall also serve as a formula for unit conversion between the rotation speed
Ω0 (in rad/s) and the command input U0 (in V).
To simulate the time response of the open-loop system, the parameters of the rotor system
are given in Table 9.2.1. We let the desired rotation speed of the propeller be
Ω0 ¼ 12000 rpm ¼ 400π rad=s. By the following MATLAB commands:
% Rotor parameters
Ka = 10; Rf = 1; Lf = 0.002; Km = 60;
Jm = 1.2; Jp = 1.1; bm = 5; kT = 4500;

% Desired propeller speed and reference input

Omega0 = 12000; % Desired speed in RPM
W0 = 1200/60*(2*pi); % Conversion from RPM to rad/s
U0 = W0*Rf*bm/Ka/Km; % Reference input U0, by Eq. (9.2.4)

Table 9.2.1 Parameters of the rotor system

Parameter Symbol Value

Polar moment of inertia of DC motor Jm 1.2 kg·m2

Damping coefficient of the motor bearing b 5 N·m·s
Field resistance Rf 1Ω
Field inductance Lf 0.002 H
Motor torque constant Km 60 N·m/A
Amplifier gain Ka 10
Torsion spring coefficient of the shaft kT 4500 N·m
Polar moment of inertia of the propeller Jp 1.1 kg·m2
690 9 Application Problems

1400

1200

1000

Speed, ω p (rpm)
800

600

400

200

0
0 0.5 1 1.5 2 2.5 3 3.5 4
Time, t (seconds)

Figure 9.2.3 Rotation speed of the propeller versus time

% Time response
P = [Lf Rf];
Q = [Jm*Jp bm*Jp kT*(Jm+Jp) bm*kT];
den = conv(P,Q);
num = kT*Ka*Km*U0/(2*pi)*60; % Conversion from rad/s to RPM
rotor_tf = tf(num,den);
step(rotor_tf);

the time response (propeller speed) of the rotor system under the input given by Eq. (9.2.4) is
plotted against time in Figure 9.2.3, where ωp ¼ θ_ p . Note that the simulation results on the
propeller speed are presented in rpm, for which the conversion between rpm and rad/s has
been implemented in a MATLAB command.
The curve in Figure 9.2.3 shows wiggles, which are due to the elasticity ðkT Þ of the shaft
connecting the motor and propeller. The elastic effect of the shaft on the system response is
further illustrated in Figure 9.2.4, with kT ¼ 700, 2000, and 8000 N·m. As observed from the
figure, a larger spring coefficient kT renders a response curve with smaller wiggles.

Parameter Variations
In operation, the motor of the rotor system may experience parameter variations, which can
be caused by thermal effects, external loads, manufacturing errors, and uncertainties in the
system modeling. Mathematically, a parameter p with variation Δp can be expressed by
p þ Δp, where Δp usually varies in a range but is unknown. Thus, if the parameters Km , Rf ,
 9.2 Speed Control of a Coupled Engine–Propeller System 691

KT = 700
15000

ω p (rpm)
10000

5000

0
0 0.5 1 1.5 2 2.5 3 3.5 4
KT = 2000
15000
ω p (rpm)

10000

5000

0
0 0.5 1 1.5 2 2.5 3 3.5 4
KT = 8000
15000
ω p (rpm)

10000

5000

0
0 0.5 1 1.5 2 2.5 3 3.5 4
t (seconds)

Figure 9.2.4 Rotation speed of the propeller versus time, with kT = 700 N·m, 2000 N·m, and 8000 N·m

and b have constant variations, the actual steady-state rotation speed of the propeller, under
the command input U0 predicted by the ideal case, will be

Ka ðKm þ ΔKm Þ Ka ðKm þ ΔKm Þ Rf b

ðωp;ss Þactual ¼ U0 ¼ Ω0
ðRf þ ΔRf Þðb þ ΔbÞ ðRf þ ΔRf Þðb þ ΔbÞ Ka Km
(9.2.5)
ð1 þ ΔKm =Km Þ ð1 þ εKm Þ
¼ Ω0 ¼ Ω0
ð1 þ ΔRf =Rf Þð1 þ Δb=bÞ ð1 þ εRf Þð1 þ εb Þ

where εp ¼ Δp=p is the percentage variation of parameter p. Obviously, the actual steady-
state speed is not the same as Ω0 in most cases unless ð1 þ εKm Þ ¼ ð1 þ εRf Þð1 þ εb Þ. For
instance, for εKm ¼ 1%; εRf ¼ 1%; and εb ¼ 2%, Eq. (9.2.5) gives

ð1  0:01Þ
ðωp;ss Þactual ¼ Ω0 ¼ 96:1%  Ω0
ð1 þ 0:01Þð1 þ 0:02Þ

For Ω0 ¼ 12 000 rpm, ðωp;ss Þactual ¼ 11532 rpm; see Figure 9.2.5 (with kT ¼ 4500 N  m).
Thus, the open-loop system in general cannot eliminate or reduce the effects of parameter
variations. This disadvantage of open-loop control is also discussed in Section 8.2.
692 9 Application Problems

12000

10000
11532 rpm

8000

Speed, ω p (rpm)
6000

4000

2000

0
0 0.5 1 1.5 2 2.5 3 3.5 4
Time, t (seconds)

Figure 9.2.5 Rotation speed of the propeller versus time, for Ω0 ¼ 12000 rpm and with parameter
variations: εKm ¼ 1%; εRf ¼ 1%; and εb ¼ 2%

Effect of Nonlinear Air Damping

If the drag torque bD θ_ p sgnðθ_ p Þ in Eq. (9.2.1d) cannot be ignored, as in the applications of
2

helicopters and unmanned aerial vehicles, the rotor system becomes nonlinear. In this case,
transfer function formulations and block diagrams in the s-domain are invalid for modeling
and simulation. However, state-space representations and block diagrams in the time
domain are applicable to this system.
In this subsection, we shall use a time-domain block diagram as a tool for modeling and
simulation of the nonlinear rotor system. To this end, we rewrite Eqs. (9.2.1b)–(9.2.1d) as
follows

1 
Dif ¼ v m  Rf i f
Lf
1
D2 θm ¼ ðτm  bDθm  τs Þ (9.2.6)
Jm
1
D2 θp ¼ ðτs  τD Þ
Jp
d
where D ¼ , vm ¼ Ka u, and τs and τD are the spring torque and drag torque given by
dt
τs ¼ kT ðθm  θp Þ and τD ¼ bD θ_ p sgnðθ_ p Þ.
2

As in the linear case, the desired propeller speed is Ω0 ¼ 12000 rpm, and the command
input is uðtÞ ¼ U0 , with Ω0 and U0 related by Eq. (9.2.4). Following the steps presented in
Section 6.5, a time-domain block diagram of the rotor system is obtained in Figure 9.2.6.
 9.2 Speed Control of a Coupled Engine–Propeller System 693

∫
θm θm θm
∫ ∫
u vm + if τm + +
Ka 1/Lf Km 1/Jm
− − −
Input

Rf b

+
τs
KT
−

sgn + θp θp θp
X bD
τD
−
1/Jp ∫ ∫
θp

Output

Figure 9.2.6 Time-domain block diagram of the rotor system with a nonlinear air-damping torque

2*pi/60 1*5/10/60 10 +− 1/(0.002) 1 60 +− +− 1/1.2 1 1

u Vm s i_f Tm s s
Input (RPM) Ka Km
RPM to rad/s 1/Lf 1/Jm
b*Rf/Ka/Km
Rf 5 theta_m
d(theta_m)/dt
1 b

T_s
4500 −+
k_T
T_s
theta_p
1 d(theta_p)/dt 1
+− 1/1.1 s s
1/Jp

T_D 60/(2*pi)

0.002 output
rad/s to RPM
d(theta_p)/dt
b_D

Figure 9.2.7 A Simulink model of the rotor system with a nonlinear air-damping torque

With the time-domain block diagram, a Simulink model of the open-loop system is easily
built; see Figure 9.2.7, where the input is the desired rotation speed (1200 rpm), the output
shown in the scope block is the propeller speed in rpm, and the gain Rf b=ðKa Km Þ is the
conversion factor given in Eq. (9.2.4). To present the propeller speed in rpm, two gains for
conversion between rpm and rad/s, 2π=60 and 60=ð2πÞ, are installed in the Simulink model.
For simplicity in the subsequent analysis, the blocks from u to θ_ p in Figure 9.2.7 can be
grouped into a subsystem; see Figure 9.2.8, where the subsystem is named Rotor. To see or to
edit the subsystem, simply double-click the box of the subsystem. The block diagrams in
Figures 9.2.7 and 9.2.8 are equivalent.
694 9 Application Problems

2*pi/60 1*5/10/60 u d(theta_p)/dt 60/(2*pi)

u d(theta_p)/dt
Input (RPM) Output
RPM to rad/s rad/s to RPM
b*Rf/Ka/Km Rotor

Figure 9.2.8 An equivalent Simulink subsystem model of the rotor system

12000

10000

8000
ω p (rpm)

6000

4000 bD = 0
bD = 0.0005
2000 bD = 0.001
bD = 0.002

0
0 0.5 1 1.5 2 2.5 3 3.5
t (s)

Figure 9.2.9 Rotation speed of the propeller versus time, without feedback control: bD = 0, 0.0005, 0.001,
and 0.002 N·m·s2

In a numerical simulation, the following four cases of air-damping coefficient are con-
sidered: bD ¼ 0, 0.0005, 0.001, and 0.002 N·m·s2. Here, the inclusion of the case of bD ¼ 0 is
two-fold: to validate the Simulink model with the linear model given by Eq. (9.2.2), and to
investigate the effect of air damping on the system response. In all four cases, the system
parameters in Table 9.2.1 are used.
Figure 9.2.9 shows the time histories of the propeller rotation speed ðωp ¼ θ_ p Þ in the four
air-damping cases, which are plotted by running the Simulink model in Figure 9.2.7. It can be
seen that the final (steady-state) value of the propeller speed decreases as the air-damping
coefficient increases. Therefore, under the influence of the drag torque, the open-loop
system, which is designed based on the ideal case ðbD ¼ 0Þ, cannot reach the desired rotation
speed (12000 rpm). Note that the curve with bD ¼ 0 (no air damping) is the same as that
obtained by the transfer function model in Figure 9.2.3.
Because the air-damping coefficient bD is difficult to predict beforehand, it is not efficient to
adjust the command input u in real time for regulating the propeller speed in this open-loop
 9.2 Speed Control of a Coupled Engine–Propeller System 695

configuration. This disadvantage of open-loop control, along with the previously mentioned
issue of parameter variations, necessitates feedback control of the rotor system.

9.2.2 Closed-Loop System

To deal with the effects of parameter variations and nonlinear air damping, feedback control
of the rotor system is installed in Simulink. Figure 9.2.10 shows a Simulink model of a closed-
loop system that consists of the subsystem Rotor in Figure 9.2.8 and a PID controller block.
It is assumed that a noncontact speed sensor is installed to measure the rotation speed of the
propeller shaft (not shown in the figure). The PID controller block implements the following
PID control algorithm with a low-pass filter on D (derivative) action:

1 N d
KP þ KI þ KD ; D¼ (9.2.7)
D 1 þ N=D dt

where N is a filter coefficient. Theoretically, as N goes to infinity, the third term in the
expression (9.2.7) becomes classic D action, KD D. Refer to Section 8.3 for the basic concepts
of PID control. Again, to show the simulation results in rpm, two gain blocks for conversion
between rpm and rad/s are installed in the Simulink model.
In the simulation using the model in Figure 9.2.10, the parameters of the rotor listed in
Table 9.2.1 are used and the controller parameters are chosen as follows: KP ¼ 0:7, KI ¼ 4,
KD ¼ 0:0007, and N ¼ 200. The propeller speed of the controlled rotor is plotted against time
in Figure 9.2.11, for four different cases of air damping: bD ¼ 0:0005, 0.001, 0.002, and
0.005 N·m·s2. Comparison of Figures 9.2.9 and 9.2.11 shows that the PID controller
can effectively regulate the propeller speed to the reference input (desired speed)
Ω0 ¼ 12000 rpm. In the case of bD ¼ 0:005 Nms2 , which describes a relatively large drag
torque, running the Simulink model in Figure 9.2.7 gives a steady-state propeller speed of the
open-loop system around 6947 rpm, which is much lower than the desired 12 000 rpm.
However, under the feedback controller, the propeller speed eventually reaches the desired
speed, as shown in Figure 9.2.11(d).
In the above-mentioned four cases of air damping, the PID control gains are the same.
This indicates that even if the air damping is unpredictable, the PID feedback controller is
able to maintain the speed of the propeller as desired.

2*pi/60 +− 1*5/10/60 PID(s) u d(theta_p)/dt 60/(2*pi)

e u d(theta_p)/dt
Input (RPM) output
RPM to rad/s rad/s to RPM
b*Rf/Ka/Km Rotor

Figure 9.2.10 Simulink model of a closed-loop system

696 9 Application Problems

bD = 0.0005 bD = 0.001
14000 14000

12000 12000

10000 10000
ω p (rpm)

ωp (rpm)
8000 8000

6000 6000

4000 4000

2000 2000

0 0
0 1 2 3 4 5 0 1 2 3 4 5
t (s) t (s)
(a) (b)

bD = 0.002 bD = 0.005
14000 14000

12000 12000

10000 10000
ω p (rpm)

ω p (rpm)
8000 8000

6000 6000

4000 4000

2000 2000

0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
t (s) t (s)
(c) (d)

Figure 9.2.11 Time histories of the propeller speed, with PID control: (a) bD ¼ 0:0005 Nms2 ;
(b) bD ¼ 0:001 Nms2 ; (c) bD ¼ 0:002 Nms2 , (d) bD ¼ 0:005 Nms2

Finally, a rotor system with the combined effects of air damping and parameter variations
is simulated. The damping coefficient bD ¼ 0:002 Nms2 and the percentage parameter
variations defined in Eq. (9.2.5) are εKm ¼ 0, εRf ¼ 5%, and εb ¼ 5%. In the closed-loop
system, the same PID control gains (KP ¼ 0:7, KI ¼ 4, KD ¼ 0:0007, and N ¼ 200) are
assigned. Figure 9.2.12 shows the time histories of the propeller rotation speed, without
and with feedback control. It can be seen that the propeller speed without control is
significantly reduced to 8201 rpm. Under the PID controller, the propeller speed reaches
the desired value quickly (in about 2.5 seconds).
In summary of this section, the rotation-speed regulation of a coupled motor–propeller
system via PID feedback control has been presented, and the utility of MATLAB/Simulink
in the relevant modeling and simulation has been demonstrated. Because the rotor system in
consideration has nonlinearity due to air damping, the gains of a feedback controller depend
on the reference input (the desired rotation speed). One practical way to deal with different
 9.3 Modeling and Analysis of a Thermomechanical System 697

12000

10000

8000
ω p (rpm)
6000

4000

2000 No control
PID control
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
t (s)

Figure 9.2.12 Time histories of the propeller speed of the rotor with air damping (bD ¼ 0:002 Nms2 ) and
parameter variations (εKm ¼ 0, εRf ¼ 5% and εb ¼ 5%): dashed line – no control; solid line – PID control

rotation speeds in operation is to build up a look-up table for different scenarios, which can
also be conveniently done with MATLAB/Simulink.

9.3 Modeling and Analysis of a Thermomechanical System (a Bimetallic

Strip Thermometer)
A bimetallic strip thermometer is a cantilever beam created by bonding together two layers
of metals with different thermal expansion coefficients. A typical pair of metals is brass and
steel, with thermal expansion coefficients of approximately 18.7 and 12 ppm/°C (parts
per million per degree Celsius). The bonding is done along the length of the metal strip,
and is achieved by either fusing them together or joining them mechanically with rivets.
When heated, the two metals expand differently, thus making the cantilever beam bend in the
direction of the lesser coefficient of thermal expansion. The deflection of the free end of
a bimetallic strip is proportional to the temperature change, making this device one of the
most durable thermometers. To improve its performance, the cantilever bimetallic beam is
made very long and formed into a spiral as shown in Figure 9.3.1.
Some applications of this technology include the popular meat thermometers, analog
thermostats, electrical breakers, and many others.
698 9 Application Problems

4 5 6 7
3 8
2
9
1

10
0
Steel

Brass or copper

Figure 9.3.1 Bimetallic spiral strip thermometer

9.3.1 Dynamic Modeling of a Simple Bimetallic Strip

Consider a cantilevered bimetallic strip as shown in Figure 9.3.2(a).
With increasing temperature, the metal layers of this cantilever beam expand. Since the
thermal expansion coefficient of the top copper layer is larger than that of the bottom steel
layer, the beam bends downward. The radius of curvature of the deformed beam is related to
the temperature increase as follows:

h
R¼ (9.3.1)
ðα1  α2 ÞΔT

where h is the thickness of the bimetallic strip, α1 and α2 are the thermal expansion coeffi-
cients, and ΔT is the temperature increase.
As stated in the literature on mechanics of materials, the inverse of the radius of curvature
equals the second derivative of the free-end deflection of the cantilever beam (x is measured
from the clamped end of the cantilever beam):

1 d2 zðxÞ
¼ (9.3.2)
R dx2

Substituting the expression for R (Eq. (9.3.1)) into Eq. (9.3.2) and integrating twice, the
expressions for the slope of the bending beam axis are obtained:

dzðxÞ ðα1  α2 ÞΔT

¼ x þ C1 (9.3.3)
dx h

and for the deflection of the beam free end:

ðα1  α2 ÞΔT 2
zðxÞ ¼ x þ C 1 x þ C2 (9.3.4)
2h
 9.3 Modeling and Analysis of a Thermomechanical System 699

fe
w z
L h1 m2
x h2 m1
α1
z(x)
h
k1 k2 c
α2 z
R

(a) (b)

Figure 9.3.2 Bimetallic strip thermometer: (a) schematic; (b) lumped-parameter model

At the clamped end ðx ¼ 0Þ both the slope and deflection are zero; thus, both integration
constants C1 ¼ C2 ¼ 0 are derived.
Maximum deflection is at the free end of the cantilever beam ðx ¼ LÞ, and it is obtained
from Eq. (9.3.4) as

ðα1  α2 ÞL2
z max ¼ zðLÞ ¼ ΔT (9.3.5)
2h

Equation (9.3.5) constitutes a static mathematical model of a bimetallic strip thermom-

eter. It shows that the measured maximum deflection is proportional to the temperature
ðα1  α2 ÞL2
change, where the constant coefficient depends on the geometry and materials of
2h
the thermometer.
A bimetallic strip thermometer is a dynamic system. When its cantilever beam is subject to
a temperature change, it starts bending and continues to do so until settling into a steady state.
The steady-state value of the free-end deflection is the parameter z max , derived above. To analyze
the dynamic system and ascertain its convergence to a steady state for a constant value of ΔT,
when a reading is taken, we create a lumped-parameter model as shown in Figure 9.3.2(b).
Two metal strips are represented by two collocated masses m1 and m2 , supported by springs
in a parallel configuration (these springs represent elastic properties of the strips). The effect
of temperature change is represented by the action of a virtual force fe : applied to the masses,
it causes the same displacement as zðLÞ. Energy losses in this lumped-parameter model are
represented by a viscous damper with the damping coefficient c.
Considering that each strip is a thin parallelepiped with length L, width w, height h1 and h2 ,
respectively, and density ρ1 and ρ2 , respectively, the equivalent mass and stiffness are
computed according to the tables in Chapter 3 and the literature on the strength of materials.
wh1 3
The moments of inertia of each strip compiling the cantilever beam are I1 ¼ and
12
wh2 3
I2 ¼ , respectively.
12
700 9 Application Problems

33
m ¼ m1 þ m2 ¼ wLðρ1 h1 þ ρ2 h2 Þ
140
(9.3.6)
3 w  
k ¼ k1 þ k2 ¼ 3 ðE1 I1 þ E2 I2 Þ ¼ 3 E1 h1 3 þ E2 h2 3
L 4L

where E1 and E2 indicate the elasticity modulus (Young’s modulus) for each strip.
At steady state (equilibrium), the assumed virtual force fe equals the spring force, which is
computed using Eqs. (9.3.5) and (9.3.6) as follows:
 
w  
3 ðα1  α2 ÞL
2 wðα1  α2 Þ E1 h1 3 þ E2 h2 3
fe ¼ kzðLÞ ¼ 3 E1 h1 þ E2 h2
3
ΔT ¼ ΔT (9.3.7)
4L 2h 8Lðh1 þ h2 Þ

The thermometer coefficient,

 
wðα1  α2 Þ E1 h1 3 þ E2 h2 3
KT ¼
8Lðh1 þ h2 Þ

is a constant that depends on the design of the given thermometer (its geometry and
materials).
When the temperature change is a function of time, the displacement of the free end of the
cantilever bimetallic beam is also a function of time, and, consequently, the virtual force
becomes fe ¼ kzðtÞ ¼ KT ΔTðtÞ. Using Newton’s second law to derive the governing equation
for this dynamic system (see Chapter 3 for detailed discussion on such derivations) we obtain

z ¼ fe  c_z  kz ) m€
m€ z þ c_z þ kz ¼ KT ΔTðtÞ (9.3.8)

Since this is a linear time-invariant second-order system, a transfer function model can be
used for generating a response. Taking the Laplace transform of Eq. (9.3.8) the following
transfer function is derived, considering that the temperature change ΔTðtÞ is the system
input, and the deflection of the free end of the cantilever bimetallic beam zðtÞ is the output:

ZðsÞ KT
¼ (9.3.9)
ΔTðsÞ ms2 þ cs þ k

The poles of this system are

c 1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p1;2 ¼   c2  4km
2m 2m

They may be real or complex conjugates, depending on the values of k, c, and m, but since the
system is built in such a way that c2  4km ≠ 0, in all cases the real parts of these poles are
strictly negative. That means that this system is stable (see Chapter 7 for a discussion on
stability). Therefore, the final-value theorem can be used to find the expected steady-state
value, assuming that this system is subject to a step input ΔTðsÞ ¼ T=s:
 9.3 Modeling and Analysis of a Thermomechanical System 701

  
KT T TKT
zss ¼ lim sZðsÞ ¼ lim s  ¼ (9.3.10)
s→0 s→0 ms2 þ cs þ k s k

9.3.2 Numerical Simulations: Comparison of Two Bimetallic Strip Thermometers

Given the numerical values of the system parameters, let us compare two thermometers with
bimetallic strips of the same size but different material pairs – copper–steel and aluminum–
tungsten. Both thermometers are subject to a temperature increase ΔT = 20 °C.

Copper–steel thermometer:

Strip 1 (copper): α1 ¼ 17:8  106 deg1 ; E1 ¼ 117  109 N=m2 ; ρ1 ¼ 8300 kg=m3 ,
Strip 2 (stainless steel): α2 ¼ 10:8  106 deg1 ; E2 ¼ 180  109 N=m2 ; ρ2 ¼ 7480 kg=m3

Aluminum–tungsten thermometer:

Strip 1 (aluminum): α1 ¼ 21:3  106 deg1 ; E1 ¼ 69  109 N=m2 ; ρ1 ¼ 2700 kg=m3 ,

Strip 2 (tungsten): α2 ¼ 4:5  106 deg1 ; E2 ¼ 400  109 N=m2 ; ρ2 ¼ 19300 kg=m3

The common geometry and viscous damping for both thermometers is as follows:
w ¼ 50  106 m; L ¼ 350  106 m; c ¼ 2:3  107 Nsec=m; h1 ¼ 100  109 m;
h2 ¼ 1  106 m
The derived mathematical model is implemented in Mathematica using the functions
TransferFunctionModel and OutputResponse (see Figure 9.3.4 for the code); the
generated dynamic response of both thermometers is shown in Figure 9.3.3.
As seen in Figure 9.3.3, the copper–steel thermometer produces a reading (settles into
a steady state) much faster than the aluminum–tungsten one. Additionally, these two
thermometers need to have different scales due to the magnitude of their respective steady-
state readings.
Computing the model parameters (damping ratio and natural frequency), as detailed in
Chapter 7, allows us to find the settling time (Eq. (7.1.70)) and the percentage overshoot (Eq.
(7.1.65)) for both thermometers analytically. These results are presented in Table 9.3.1.

Table 9.3.1 Numerical results for the comparison of the two thermometers

ξ ωn , rad/s ts , s MP ; % zðLÞ, m
Copper–steel 0.086 39139 0.0012 76.3 7:8  106
Aluminum–tungsten 0.037 38011 0.0028 88.9 18:7  106
702 9 Application Problems

Figure 9.3.3 Dynamic response of two bimetallic strip thermometers: copper-steel and aluminum-
tungsten

While different scales and a shorter settling time do not necessarily make one bimetallic
strip thermometer better than the other (the choice of device depends on the intended
application), the greater overshoot may render the aluminum–tungsten thermometer less
desirable. A greater overshoot coupled with the larger steady-state value means that, while
the bimetallic strip is bending, the pointer will experience significant rotation (see
Figure 9.3.1), which will strain the device and may cause damage.

9.4 Modeling and Analysis of an Electro-Thermo-Mechanical System

(a Resistive-Heating Element)
Resistive-heating elements are used in many devices and industrial processes. An electric
fuse, which melts due to Joule effect and breaks the circuit when the flowing current exceeds
the designated threshold value, is a commonly used safety device. Electronic cigarettes
vaporize propylene glycol and vegetable glycerin by utilizing the heating filaments.
Common appliances and tools such as electric stoves, electric space heaters, cartridge
heaters, hair dryers, and soldering irons, just to name a few, contain resistive-heating
elements to fulfill their intended functions. The simulation and analysis of the problem
presented in this section gives the reader an insight into the operation of such devices.
 9.4 Modeling and Analysis of an Electro-Thermo-mechanical System 703

Figure 9.3.4 Generating dynamic system response: Mathematica code

704 9 Application Problems

9.4.1 System Description

A stainless-steel heating element is represented by a deformable solid cylinder of radius r and
length L0; it is treated as a resistor with initial resistance R0. The element is heated by means
of an external electrical circuit as shown in Figure 9.4.1. The applied voltage v generates heat
due to the Joule effect and, subsequently, elongation of the heating element.
The heating element is immersed in a fluid, which is at constant temperature T∞ ; heat loss
occurs through convection with the convective coefficient h. The surface temperature of the
heating element is Ts .
The known properties of the heating element are: mass m, mass density ρ, specific heat c,
electrical resistivity ρel , thermal coefficient of resistance αR , and linear coefficient of thermal
expansion α.
The following assumptions can be made to simplify the derivations without losing the
expressiveness of the model:

(a) The change of the cross-sectional area and dimensions with the temperature is negligible.
(b) The change of the mass density with the temperature is negligible.
(c) The electrical resistivity is constant, and the electrical resistance is linearly increasing
with the change of temperature.
(d) The electrical-to-mechanical energy conversion is instantaneous.
(e) One end of the heating element is fixed; elongation is axial and occurs only at the free
end.

9.4.2 Dynamic Modeling

As described in Chapter 6, for every combined system the initial step in modeling involves the
derivation of the governing equations for every subsystem.

qout
T∞
Fixed end Elongating end
qin r
Ts x

+ −

v(t)
L0

Figure 9.4.1 Heating element model

 9.4 Modeling and Analysis of an Electro-Thermo-mechanical System 705

Consider the thermal subsystem first. The governing equation for such a system is derived
in Chapter 5 as follows (Eq. (5.2.3)):

dT
C ¼ qin  qout (9.4.1)
dt

where C is the thermal capacitance of the heating element, T describes the temperature of the
heating element, and qin  qout is the net heat flow. Since the specific heat of the given heating
element is not a function of temperature, and the mass is assumed to be constant, the thermal
capacitance is defined using Eq. (5.2.2) as

C ¼ mc ¼ ρVc ¼ ρðπr2 L0 Þ c (9.4.2)

where m is mass of the heating element, c is the specific heat, ρ is the mass density, V is the
volume, r is the radius, and L0 is the initial length.
The output heat flow represents heat loss through convection, as derived in Eq. (5.1.3).
The expression for qout must also consider two important factors: (a) the surface area
involved in this heat transfer is the lateral area of the heating element; (b) the elongation
of the heating element is one-dimensional along the x-axis:
 
qout ¼ hð2πrLÞðTs  T∞ Þ ¼ 2πrhL0 1 þ αðTs  T∞ Þ ðTs  T∞ Þ (9.4.3)

where h is the convective coefficient, α is the linear coefficient of thermal expansion, Ts is the
surface temperature of the heating element, and T∞ is the temperature of the fluid at some
distance away from the surface.
The input heat flow occurs because of the Joule effect – the electrical energy dissipated by
the resistor that models the given heating element is transformed into heat. Assuming lossless
wires, the coupling equation describing the relationship between the thermal and electrical
subsystems is derived as follows:

qin ¼ P (9.4.4)

where qin is the input heat flow rate, and P is the electrical power absorbed by the heating
element and converted into heat.
Since the given heating element is modeled as a linear resistor, the electrical power
dissipated by it as heat is described with Eq. (4.1.11). Then the input heat flow rate becomes

v2
qin ¼ P ¼ (9.4.5)
R

where v is the applied voltage, and R is the resistance of the heating element at the tempera-
ture Ts .
706 9 Application Problems

Considering that the electrical resistance of the heating element linearly increases with the
change of temperature, this resistance at the temperature Ts can be expressed in terms of the
initial resistance as follows:
 
R ¼ R0 þ αR R0 ðTs  T∞ Þ ¼ R0 1 þ αR ðTs  T∞ Þ (9.4.6)

The initial resistance is computed with Eq. (4.1.10) as follows:

ρel L0
R0 ¼ (9.4.7)
πr2

Substituting Eqs. (9.4.6) and (9.4.7) into Eq. (9.4.5) the expression for the input heat flow is
obtained:

πr2
qin ¼ v2   (9.4.8)
ρel L0 1 þ αR ðTs  T∞ Þ

Considering that the fluid temperature T∞ is constant, the variable temperature T that
describes surface temperature of the heating element can be written as T ¼ Ts  T∞ .
Then, substituting Eqs. (9.4.2), (9.4.3), and (9.4.8) into Eq. (9.4.1), we obtain the first
governing equation for this system:

πr2
πr2 L0 ρcT_ ¼ v2  2πrhL0 ð1 þ αT ÞT
ρel L0 ð1 þ αR T Þ

which, upon simplifying, becomes

rL0 2 ρρel c T_ þ rL0 2 ρρel c αR T_ T þ 2hL0 2 ρel αR αT 3 þ
(9.4.9)
þ 2hL0 2 ρel ðαR þ αÞT 2 þ 2hL0 2 ρel T ¼ rv2

The second governing equation describes the axial elongation of the heating element due
to the temperature change:

xðtÞ ¼ α L0 TðtÞ (9.4.10)

These two equations – Eq. (9.4.9) and Eq. (9.4.10) – constitute a mathematical model of
the given dynamic system. They are clearly coupled: when the nonlinear Eq. (9.4.9), which
describes the electro-thermal component, is solved for T(t), the mechanical component – the
deformation of the heating element – can be computed.

9.4.3 Time Response of the Heating Element: Temperature Change and Elongation
The analytical solution of a nonlinear governing equation (Eq. (9.4.9)) does not exist.
Therefore, numerical simulation needs to be performed. The following system parameters
are known:
 9.4 Modeling and Analysis of an Electro-Thermo-mechanical System 707

Geometry and density:

L0 ¼ 0:1 m; r ¼ 2 mm ¼ 2  103 m; ρ ¼ 7930 kg=m3

Thermal properties:

c ¼ 490 J=kg  deg; h ¼ 32 J=m2  sec  deg;

α ¼ 12  106 deg1 ; αR ¼ 0:001 deg1

Electrical properties:

ρel ¼ 6:9  107 Ω  m; v ¼ 0:015 V

To simplify the computations, divide the first governing equation (Eq. (9.4.9)) by the
coefficient rL0 2 ρρel c and rewrite it in the following form:

2hαR α 3 2hðαR þ αÞ 2 2h 1
T_ þ αR T_ T þ T þ T þ T¼ 2 v2
rρc rρc rρc L0 ρρel c

Note that since T ¼ Ts  T∞ , when voltage is applied and heating starts, Tð0Þ ¼ 0 (the
temperature of the heating element is the same as the temperature of the surrounding fluid).
The numerical solution is found using either the Simulink or Mathematica function
NDSolve or its variant NDSolveValue. The code for deriving the solution and generating
the graphs of temperature difference T ¼ Ts  T∞ and elongation of the heating element is
presented in Figure 9.4.2.
The generated graphs are shown in Figure 9.4.3.
Figure 9.4.3(a) shows that within approximately 800 seconds the temperature difference
between the heating element and the surrounding fluid reaches a steady state: a constant
value of approximately 1.02 degrees. The deformation of a stainless-steel heating element
also exhibits a steady-state response of approximately 1.22 μm. While a small deformation is
desirable, the achieved steady-state temperature difference is clearly insufficient for mean-
ingful heating. In order to improve the performance of the heating element and increase the
generated temperature difference, the net heat flow needs to be increased. Decreasing the
convective heat loss is not possible since it depends on the parameters of the fluid, into which
the heating element is immersed: this fluid is specified in the technical requirements and
cannot be changed. Then, designing a more efficient heating element involves increasing the
input heat flow rate qin by either requiring a higher input voltage or changing the material of
the heating element to one with smaller electrical resistivity ρel .
Changing the geometry of the heating element does not provide the desired temperature
difference. Increasing the length decreases the magnitude of the input, thus decreasing
the steady-state response. While decreasing the length to a half of the original value (setting
708 9 Application Problems

Figure 9.4.2 Generating the graphs of the temperature and elongation of the heating element:
Mathematica code
 9.4 Modeling and Analysis of an Electro-Thermo-mechanical System 709

Figure 9.4.3 Dynamic response of the heating element: (a) temperature; (b) elongation

Figure 9.4.4 Dynamic response of the heating element with modified geometry: (a) L0 = 0.05 m; (b) r =
10 mm

L0 = 0.05 m) helps to increase the input heat flow as seen in Figure 9.4.4(a), this increase is
insufficient, and making the heating element even shorter is impractical.
Increasing the radius makes the settling into a steady state significantly slower, while
providing a marginal increase in the temperature difference; see Figure 9.4.4(b).
Increasing the input voltage to 0.075 V generates a temperature difference of approxi-
mately 25.8 degrees and a deformation of 29.8 μm, as shown in Figure 9.4.5.
In many practical applications a specific temperature difference is required to be produced
by the heating element. For example, assume that the required temperature difference for the
given heating element is at least 80 °C. Numerical simulations show that this is achieved by
710 9 Application Problems

Figure 9.4.5 Dynamic response of the heating element with increased input voltage: (a) temperature,
(b) elongation

Figure 9.4.6 Dynamic response of the heating element to an input of 0.15 V: (a) temperature,
(b) elongation

applying an input voltage of approximately 0.15 V (see Figure 9.4.6(a)), which is ten times
larger than the originally intended input. The elongation of the heating element in response
to this input is still small – only 0.1 mm (see Figure 9.4.6(b)).
If a higher input voltage is unavailable, a voltage-isolation amplifier could be added to the
electrical subsystem to increase the given voltage as required (see Section 4.7.1 for detailed
information). Nonetheless, a higher voltage in the electrical circuit of the heating element
system may be undesirable for safety reasons. In this case, changing the material of a heating
element becomes necessary. If the heating element is made of copper instead of stainless steel,
 9.5 Feedback Control of a Liquid-Level System for Water Purification 711

Figure 9.4.7 Dynamic response of the copper heating element: (a) temperature; (b) elongation

a performance improvement is achieved with lower input voltage, as shown in the following
simulations.
The copper parameters are:

ρ ¼ 8900 kg=m3 ; c ¼ 385 J=kg  deg;

α ¼ 17  106 deg1 ; αR ¼ 0:0039 deg1 ; ρel ¼ 17  109 Ω  m

Note that copper has an approximately 100 times smaller electrical resistivity than
stainless steel, with the other parameters comparable in order of magnitude.
By using copper, the temperature difference for the same size of heating element and the
same input voltage (0.015 V) increases by more than 30 times (from 1.02 to 36.2 degrees), as
evidenced in Figure 9.4.7(a). Deformation also increases significantly, but it is still in microns
(approximately 61.6 μm) as seen in Figure 9.4.7(b).
Alas, the input voltage still needs to be increased to reach the objective of at least an 80 °C
temperature difference, but that increase does not have to be as drastic as it was in the case of
a stainless-steel heating element. Doubling the input voltage (to 0.03 V) yields a temperature
difference of 114.23 degrees and elongation of 0.19 mm. as seen in Figure 9.4.8.

9.5 Feedback Control of a Liquid-Level System for Water Purification

Liquid-level systems, which are assemblies of tanks, pipes, valves, pumps, and other compo-
nents, have extensive applications in various industrial processes. A detailed discussion on
these systems is presented in Chapter 5. An example of a two-tank liquid-level system,
712 9 Application Problems

Figure 9.4.8 Dynamic response of the copper heating element to an input of 0.03 V: (a) temperature;
(b) elongation

commonly used for a small-scale water purification (two-stage biochemical process), is

presented in this section.

9.5.1 System Description

The system, shown in Figure 9.5.1(a), consists of two cylindrical tanks, the circular bottom
areas of which are A1 and A2 , respectively. Raw water is pumped into the first tank for pre-
treatment at a flow rate of qin ðtÞ. A mix of aerobic microorganisms that can digest organic
matter in water resides in both tanks, producing an effluent that is relatively free from
suspended micro-organics. Treated water moves to the second tank through a large orifice
with the cross-sectional area a1 . In the second tank, the water undergoes additional treat-
ment with a different mix of aerobic bacteria, leaving through an orifice with the cross-
sectional area a2 and flowing into a clarifying vessel where the agglomerated bacterial
particles are removed. The orifice in each tank is a slit in the cylinder wall, as shown in
Figure 9.5.1(b), defined by the height d, and the angle φ of the circular sector generating the
upper and lower boundaries of the slit.
During the operation of the system, the liquid level in tank 1 is h1 ðtÞ, and in tank 2 it is
h2 ðtÞ. The discharge coefficient Cd is the same for both tanks.
The liquid level in tank 2 is crucial to the process efficiency and quality of the effluent due
to the nature of the biochemical reactions occurring in this tank. A linear controller needs to
be designed, with a piecewise-continuous reference value r defined in accordance to the
specifics of the process.
 9.5 Feedback Control of a Liquid-Level System for Water Purification 713

qin

A1

h1 a1
qm
ϕ
Tank 1 d
A2

h2 a2

qout
Tank 2
(a) (b)

Figure 9.5.1 Liquid-level system: (a) description, (b) orifice

9.5.2 Dynamic Modeling

A mathematical model of the liquid-level system needs to be derived and analyzed prior to
the controller design. As derived in Chapter 5 (Eq. (5.5.12)), the governing equations for each
tank are as follows:
8
>
> dh1
¼ qin ðtÞ  qm
< A1
dt
(9.5.1)
>
>
:A2 dh2 ¼ qm  qout
dt

where qin ðtÞ is the inflow of raw water into tank 1, qm is the outflow from tank 1 through the
orifice a1 , h1 and h2 are the liquid levels in tank 1 and tank 2, respectively, and qout is the
outflow from tank 2 through the orifice a2 .
According to Torricelli’s principle. the outflow qmax through the orifice of area a cannot
exceed aρvmax , where ρ is the density of the fluid, and vmax is its maximum speed. This
maximum speed can be derived from the law of the conservation of energy as
pffiffiffiffiffiffiffiffi
vmax ¼ 2gh, where h is the height of the fluid above the orifice, and g is the gravitational
acceleration. Hence, the outflow through the orifice cannot exceed the following value:
pffiffiffiffiffiffiffiffi
qmax ¼ aρ 2gh. The actual outflow will be less than this because of frictional effects and is
derived as
pffiffiffiffiffiffiffiffi
qactual ¼ Cd a 2gh (9.5.2)

where Cd is the discharge coefficient.

714 9 Application Problems

Applying the derived expression for the outflow through an orifice (Eq. (9.5.2)) to the
governing equations (Eq. (9.5.1)) we obtain
8 pffiffiffiffiffiffiffiffiffiffi
>
> dh1
¼ qin ðtÞ  Cd a1 2gh1
< A1
dt
(9.5.3)
>
>A dh pffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffi
: 2
2
¼ Cd a1 2gh1  Cd a2 2gh2
dt

The form of these first-order ordinary differential equations strongly resembles the form
of state equations. Defining h1 and h2 as state variables, we obtain the state-space model of
this system, considering that the output is the liquid level in tank 2:
8
> dh1 1 Cd a1 pffiffiffiffiffiffiffiffiffiffi
>
< ¼ qin ðtÞ  2gh1
dt A1 A1
>
> dh C a pffiffiffiffiffiffiffiffiffiffi C a pffiffiffiffiffiffiffiffiffiffi
: 2 ¼ d 1 2gh1  d 2 2gh2
dt A2 A2
y ¼ h2 (9.5.4)

For the stability of the biochemical reactions, the volumes of liquid in both tanks must stay
constant at all times, which means that the liquid levels must stay constant. Thus, a steady-
state response of the system is desired.
The tanks need to be filled up to a predefined liquid level for the bacterial mix to be added
and reactions to start. This moment in time is considered the starting point t ¼ 0, and liquid
levels h1 ð0Þ and h2 ð0Þ are the initial conditions which need to be incorporated into the derived
state-space model.
This nonlinear state-space model does not have an analytical solution; hence, it needs to be
solved via numerical integration using Runge–Kutta methods as described earlier in
Section 2.8. These methods are implemented in computational software packages such as
MATLAB (function ode45) and Mathematica (function NDSolve and its variants). It is also
possible to use Mathematica functions unique to state-space system representations such as
NonlinearSystemModel, StateResponse, and OutputResponse (note: the function
NonlinearSystemModel is only available for Mathematica version 12.0 or later).
A numerical simulation is run to determine whether this system reaches a steady state in
response to a constant (step) input, i.e., constant inflow of the raw water. These steady-state
values can be estimated analytically prior to the simulation in the following manner.
dh1 dh2
Assuming a step input qin ¼ quðtÞ, consider that, at steady state, ¼ 0 and ¼ 0.
dt dt
Then, the system of two algebraic equations:
8
> 1 Cd a1 pffiffiffiffiffiffiffiffiffiffi
>
<0 ¼ q  2gh1
A1 A1
(9.5.5)
>
> C a pffiffiffiffiffiffiffiffiffiffi C a pffiffiffiffiffiffiffiffiffiffi
:0 ¼ d 1 2gh1  d 2 2gh2
A2 A2
 9.5 Feedback Control of a Liquid-Level System for Water Purification 715

can be solved for the expected steady-state liquid levels:

8
>
> q2
>
<h1 ¼ 2C 2 a 2 g
d 1
(9.5.6)
>
> q2
>
:h2 ¼
2Cd 2 a2 2 g

The following system parameters are given:

qin ¼ 0:1uðtÞ m3 =sec; Cd ¼ 0:62; g ¼ 9:8 m=s2 ;

Tank 1 : r1 ¼ 0:4 m; d ¼ 3 cm ¼ 0:03 m; φ ¼ 60°; h1 ð0Þ ¼ 1:0 m;
Tank 2 : r2 ¼ 0:42 m; d ¼ 3:5 cm ¼ 0:035 m; φ ¼ 50°; h2 ð0Þ ¼ 0:5 m:

The bottom area and orifice area for each tank are computed as follows:

2πr
A ¼ πr2 and aorifice ¼ d (9.5.7)
360=φ

The Mathematica code for deriving the solution of state equations by numerical integration
and generating the graphs of liquid levels vs. time for both tanks is shown in Figure 9.5.2. Note
that the state-space model is set using the function NonlinearSystemModel, and the
solutions are generated with StateResponse, since both h1 ðtÞ and h2 ðtÞ variables are of
interest.
The output of the generated nonlinear state-space model (see Figure 9.5.3) can be verified
vs. the derived by-hand result – note the incorporation of the initial conditions.
The generated liquid-level graphs are shown in Figure 9.5.4.
Finding the numerical values for the steady-state response is straightforward with the
Mathematica algebraic and transcendental equation solver Solve. It is also possible to
approximate the settling time with the FindRoot function, considering a 0.5% allowed error
and using the graphs to estimate this settling time at approximately 400 seconds for the first
tank, and 600 seconds for the second tank. Code for these operations is shown in
Figure 9.5.5.
The results are as follows:

Steady-state liquid level: 8.41 m (tank 1) and 8.1 m (tank 2)

Settling time: 415.3 s ≌ 6.92 min (tank 1) and 602.8 s ≌ 10.05 min (tank 2)

As the simulation demonstrates, this system reaches a steady state in response to a step
input. The magnitude of the steady-state liquid levels depends on the input, as evidenced by
Eq. (9.5.6), and shown further in Figure 9.5.6.
The magnitude of the steady-state response to a specific step input may or may not be
optimal for the biochemical reactions occurring in the liquid. Additionally, environmental
716 9 Application Problems

Figure 9.5.2 Generating the graphs of liquid levels for both tanks: Mathematica code

Figure 9.5.3 Nonlinear state-space model in general form

 9.5 Feedback Control of a Liquid-Level System for Water Purification 717

Figure 9.5.4 Dynamic response of the system: liquid levels in both tanks

Figure 9.5.5 Estimating steady-state values and settling times for liquid levels in both tanks: Mathematica
code

factors and the variability of the system parameters due to manufacturing imperfections,
uncertainties in the system modeling, accumulation of microscopic bacterial slime around
the orifices, and other factors make maintaining the optimal operating conditions impossible
in the open-loop system. Hence, a feedback control system needs to be designed for
718 9 Application Problems

Figure 9.5.6 Dynamic response of the system to different inputs: (a) qin ¼ 0:05uðtÞ; (b) qin ¼ 0:2uðtÞ

Figure 9.5.7 Influence of the system parameters variability on the dynamic response

providing a consistent quality of effluent regardless of the fluctuations of the raw water
inflow, the variability of the operating requirements, and other disturbances.
The influence of the variability of the system parameters is clearly seen in Figure 9.5.7.
A mere 1% change in the values of system parameters causes a visible (approximately 4.3%)
difference in the steady-state liquid levels: 0.35 m for tank 1, and 0.34 m for tank 2.
 9.5 Feedback Control of a Liquid-Level System for Water Purification 719

The given system is nonlinear, represented by the nonlinear state-space model (Eq.
(9.5.4)), the state equations of which can be written in a general form as follows:

_
hðtÞ ¼ ΦðhðtÞ; uðtÞÞ (9.5.8)

h1 ðtÞ
where hðtÞ ¼ is a state vector, Φ is a nonlinear function, and uðtÞ is an input vector.
h2 ðtÞ
The behavior of nonlinear equations is often complicated. The analysis of the time response
of a nonlinear system and the design of an efficient control law for it require specific
approaches, the presentation of which is typically reserved for graduate-level classes;
hence, this is beyond the scope of this book.
If a nonlinear system has an equilibrium point – a point at which the system is in a steady
state, then it can be reliably approximated with a linear system, and the analysis of the system
behavior around such equilibrium point can be performed. This process is called lineariza-
tion. We assume that for some input us ðtÞ and some initial state there exists a solution of the
nonlinear system hs ðtÞ and for input u0 there exists an equilibrium point h0 . We let this input
to be slightly perturbed to become us ðtÞ þ ΔuðtÞ; and let the initial state to be also slightly
perturbed. The solution of the system with this perturbation is expressed as hs ðtÞ þ ΔhðtÞ,
where ΔhðtÞ is small in the vicinity of the equilibrium point for all values of t. Plugging this
solution into the nonlinear state equation (Eq. (9.5.8)) and using a Taylor series expansion of
the function Φ we derive the linearized state equation at the equilibrium point ðh0 ; u0 Þ:


d d

hs ðtÞ þ ΔhðtÞ ¼ Φðhs ðtÞ þ ΔhðtÞ; us ðtÞ þ ΔuðtÞÞ
¼ Φðh0 þ Δh; u0 þ ΔuÞ
dt dt 0 ðh0
; u0 Þ 1






∂Φ
∂Φ

2

∂ Φ
Δh 2 2

∂ Φ
Δu 2
¼ Φðh0 ; u0 Þ þ
Δh þ
Δu þ @ 2
þ 2
þ ...A
∂h
∂u
∂h
2! ∂u
2!
ðh0 ; u0 Þ ðh0 ; u0 Þ ðh0 ; u0 Þ ðh0 ; u0 Þ

(9.5.9)

Considering that Φðh0 ; u0 Þ is the solution of the original unperturbed state equation, i.e.
dhs ðtÞ
¼ Φðh0 ; u0 Þ, these components are canceled in Eq. (9.5.9). The higher-order terms
dt
Δh2 Δu2
(starting from and ) can be neglected since these perturbations are assumed to be small.
2! 2!
∂Φ ∂Φ
The partial derivative terms j and j represent constant matrices, denoted
∂h ðh0 ; u0 Þ ∂u ðh0 ; u0 Þ
A and B, respectively (they are called Jacobian matrices). Redefining Δh ≜ h and Δu ≜ u, the
linearized system is obtained:
h_ ¼ A:h þ B:u (9.5.10)
where
∂Φ
A¼ j
∂h ðh0 ; u0 Þ
720 9 Application Problems

and
∂Φ
B¼ j
∂u ðh0 ; u0 Þ

For the given system these Jacobian matrices are:

2   3

∂ 1 Cd a1 pffiffiffiffiffiffiffiffiffiffi ∂ 1 Cd a1 pffiffiffiffiffiffiffiffiffiffi

6 ∂h1 A1 in q ðtÞ  2gh 1 qin ðtÞ  2gh1 7


6 A1 ∂h2 A1 A1 7

A¼6   7

4 ∂ Cd a1 pffiffiffiffiffiffiffiffiffiffi Cd a2 pffiffiffiffiffiffiffiffiffiffi ∂ Cd a1 pffiffiffiffiffiffiffiffiffiffi Cd a2 pffiffiffiffiffiffiffiffiffiffi 5

2gh1  2gh2 2gh1  2gh2
ðh
∂h1 A2 A2 ∂h2 A2 A2 1e ; h2e Þ

2 3
Cd a1 g
 pffiffiffiffiffiffiffiffiffiffiffi 0
6 A1 2gh1e 7
6 7
¼6 7 (9.5.11a)
4 Cd a1 g Cd a2 g 5
pffiffiffiffiffiffiffiffiffiffiffi  pffiffiffiffiffiffiffiffiffiffiffi
A2 2gh1e A2 2gh2e
2  3

∂ 1 Cd a1 pffiffiffiffiffiffiffiffiffiffi
2 3
6 ∂q A in q ðtÞ  2gh 1 7

1
6 in 1 A 1 7
4 A1 5
B¼6  7 ¼ (9.5.11b)
4 ∂ Cd a1 pffiffiffiffiffiffiffiffiffiffi Cd a2 pffiffiffiffiffiffiffiffiffiffi 5


2gh1  2gh2
ðh ; h Þ 0
∂qin A2 A2 1e 2e

Then the state-space model of the linearized system in expanded matrix form becomes
2 3 2 Cd a1 g
3
dh1  pffiffiffiffiffiffiffiffiffiffiffi 0 " #
6 dt 7 6 6 A1 2gh1e 7
7 h
1
6 7¼6 7∙ 1
þ qin ðtÞ (9.5.12)
4dh 5 4 Cd a1 g Cd a2 g 5 h2 A 1
2
pffiffiffiffiffiffiffiffiffiffiffi  pffiffiffiffiffiffiffiffiffiffiffi 0
dt A2 2gh1e A2 2gh2e

where the equilibrium values h1e and h2e for the input q are given by Eq. (9.5.6).
Taking the Laplace transform of the linear equation (9.5.12) and recalling that the
required output is h2 ðtÞ the transfer function of this system is derived. To simplify the
derivations, denote

Cd a1 g Cd a1 g Cd a2 g
b11 ¼ pffiffiffiffiffiffiffiffiffiffiffi ; b21 ¼ pffiffiffiffiffiffiffiffiffiffiffi ; b22 ¼ pffiffiffiffiffiffiffiffiffiffiffi
A1 2gh1e A2 2gh1e A2 2gh2e

Then, the Laplace transform of the linearized state equations becomes

8
< 1
s þ b22 s þ b22
sH1 ðsÞ ¼Qin ðsÞ  b11 H1 ðsÞ 1
A ) H1 ðsÞ ¼ H2 ðsÞ ) ðs þ b11 Þ H2 ðsÞ ¼ Qin ðsÞ
:sH ðsÞ ¼ b H ðsÞ  b H ðsÞ
1
b21 b21 A1
2 21 1 22 2

H2 ðsÞ b21
GðsÞ ¼ ¼
Qin ðsÞ A1 ðs þ b11 Þðs þ b22 Þ
 9.5 Feedback Control of a Liquid-Level System for Water Purification 721

Substituting the above values for the coefficients b11 ; b21 ; b22 and steady-state values h1e and
h2e , the following transfer function is obtained:

H2 ðsÞ Cd 2 a1 2 gQ
GðsÞ ¼ ¼ (9.5.13)
Qin ðsÞ ðCd a1 g þ A1 QsÞðCd 2 a2 2 g þ A2 QsÞ
2 2

where Q is the value of the input (0.1 m3/s)) at which the steady-state values were computed.
This transfer function is used to design the proportional–integral (PI) control –
1
GPI ðsÞ ¼ KP þ KI – using the Ziegler–Nichols method. Applying the proportional control
s
KP helps to alleviate the effects of parameter variations of the given system, as well as to
reduce the output error caused by the disturbances arising from the environmental influence
on the system. While improving the system robustness the proportional control is not
sufficient to make the system response converge to the required value. Adding the integral
1
control KI improves the response by driving the steady-state error to zero. A detailed discussion
s
of the advantages and design of feedback control systems is presented in Sections 8.2 and 8.3.
In Mathematica, the generation of the system transfer function and design of PI control are
done using the functions TransferFunctionModel, which automatically linearizes the
derived nonlinear state-space model, and PIDTune, which generates the KP and KI gains (see
Section 8.3.2 for an explanation of Ziegler–Nichols gain tuning). This code is shown in
Figure 9.5.8.
The derived control gains are as follows: KI ¼ 7:95  105 ; KP ¼ 0:013.
To visualize the performance of the controlled system, we use the operational parameter
(reference input) defined by a simple linear function h2_reference ¼ 6 m 8 t > 0. The closed-loop
system model is generated with the Mathematica functions SystemsModelSeriesConnect

Figure 9.5.8 Designing PI control using transfer functions of linearized system: Mathematica code
722 9 Application Problems

Figure 9.5.9 Dynamic response of the nonlinear system under PI control to a reference input of 6 m

and SystemsModelFeedbackConnect. The system response is then generated with the

function OutputResponse and plotted together with the reference input (see Figure 9.5.9).
As this simulation demonstrates, PI control successfully makes the system follow the
required operational parameter. If the settling time is unacceptably long, a differential
component can be added to make the controlled system faster.
The designed PI control is capable of following the operational parameter defined by
a piecewise-continuous function, such as
8
> 7:2 m 0 ≤ t ≤ 800 s
>
>9:3 m 800 < t ≤ 1200 s
<
h2_reference ¼ 8:1 m 1200 < t ≤ 1800 s
>
>
:5:9 m 1800 < t ≤ 2500 s
>
3:5 m t > 2500 s

In this case, convergence will be more difficult to achieve, but the controlled system will still
try to follow the reference trajectory as closely as possible (see the dynamic response graph in
Figure 9.5.10). The Mathematica code for this simulation is shown in Figure 9.5.11.
The validity of the original assumption that this system can be reliably approximated with
a linearized model can be illustrated with a comparison of the performance of two systems, both
controlled with the same PI control: nonlinear and linearized. As seen in Figure 9.5.12, for all but
one segment of the reference trajectory both systems exhibit nearly identical behavior.
 9.6 Sensors, Electroacoustic, and Piezoelectric Devices 723

Figure 9.5.10 Dynamic response of the nonlinear system under PI control to a piecewise-continuous
reference input

The code for this simulation is shown in Figure 9.5.13 (note the different options used in
PIDTune).

9.6 Sensors, Electroacoustic, and Piezoelectric Devices

9.6.1 Sensors
Sensors are pervasive in nature and in all fields of human activity. The generic definition of
a sensor as given by the Merriam-Webster Dictionary is “a device that responds to a physical
stimulus (as heat, light, sound, pressure, magnetism, or a particular motion) and transmits
a resulting impulse (as for measurement or operating a control).” Sensors that are most
commonly encountered in engineering systems detect a change in a physical parameter such
as, for example, acceleration, heat flux, or pressure, and transform it into an electrical signal.
Sensors typically include a sensing element and a transducer – circuitry that performs energy
conversion from one form to another, hence producing the signal suitable for processing.
Obviously, since sensor operation involves energy conversion, such a device is
a compound dynamic system. An example of a sensor that is an electromechanical system
is a galvanometer (see Section 6.3.2), where the change in electrical current is converted to
mechanical energy (rotation of a pointer) that is subsequently measured.
724 9 Application Problems

Figure 9.5.11 Applying PI control and plotting the system response to a piecewise-continuous reference
input: Mathematica code

Other electromechanical sensing devices of interest are strain gauges and accelerometers.
A strain gauge, shown in Figure 9.6.1, is used to measure deflection, stress, and pressure.
A metallic strain gauge consists of a metallic foil or a fine wire, arranged in a grid pattern that
is bonded to a thin backing (carrier). The carrier is attached to an element, the strain of which
needs to be measured. The gauge bends together with the tested element, and the length of
the gauge grid changes. This elongation translates into a linear change in the gauge electrical
resistance. The device’s electrical subsystem consists of a specific circuit called the
Wheatstone bridge, which is used to measure the electrical resistance of the gauge terminals.
 9.6 Sensors, Electroacoustic, and Piezoelectric Devices 725

Figure 9.5.12 Dynamic response of the nonlinear and linear systems under PI control to a piecewise-
continuous reference input

Figure 9.5.13 Applying PI control and plotting the nonlinear and linearized system response:
Mathematica code
726 9 Application Problems

Electric circuit

Gauge
width

Terminals
Carrier Gauge length

Figure 9.6.1 Strain gauge

Example 9.6.1 Displacement sensor to detect a sinusoidal motion

Consider a displacement sensor for the detection of a sinusoidal motion. The sensor uses
a microcantilever beam, and four strain gauges connected in a Wheatstone bridge (see
Figure 9.6.2). The sinusoidal motion of the supporting structure (the motion that needs to be
detected) generates oscillatory bending of the microcantilever beam in the shown direction of
u (see Figure 9.6.2(a)). Two strain gauges are installed on each face of the microcantilever beam
(face 1 is shown in Figure 9.6.2(a), face 2 is the opposite one) at its root.
Modeling of this electromechanical system starts with the formulation of the governing equa-
tions for each subsystem separately, proceeds with the derivation of the relationship between the
axial strain and the input/output voltages of the Wheatstone bridge, and finally synthesizes all the
information into a coherent system model that relates the sensor sensitivity K to the sinusoidal
input frequency ω. The last step encompasses a numerical simulation, the generation of the graph
of K vs. ω, and analysis of the results.

Figure 9.6.2 Displacement sensor: (a) strain gauges and microcantilever-beam assembly;
(b) Wheatstone bridge; (c) lumped-parameter model
 9.6 Sensors, Electroacoustic, and Piezoelectric Devices 727

The sensitivity of a strain-gauge displacement sensor is related to the variation in the

Wheatstone-bridge output voltage as follows:

Δv0 ¼ Kx (9.6.1)

where K is sensitivity and x is the variation of the mechanical parameter that is being
monitored. For this problem, x is the relative displacement of the free end of the micro-
cantilever beam with respect to the oscillating supporting structure.
A lumped-parameter model of the microcantilever beam and strain gauges assembly is
shown in Figure 9.6.2(c). In this model, m and k are the effective mass and effective stiffness
(refer to Chapter 3 for a detailed discussion on finding the effective mass and stiffness for
a continuous medium such as a beam). The governing equation for this mechanical subsys-
tem is derived using Newton’s second law:

z ¼ kðz  uÞ
m€ (9.6.2)

where u is the oscillatory displacement of the supporting structure, and z is the displacement
of the free end of the microcantilever beam.
Denoting the relative displacement of this free end as x ¼ z  u, the governing equation is
rewritten as follows:

d2
m ðx þ uÞ ¼ kx ) m€
x þ kx ¼ m€
u (9.6.3)
dt2

Since the input u is a periodic function, the frequency response needs to be derived (refer to
Chapter 7 for details). Taking the Laplace transform of Eq (9.6.3):

X ðsÞ ms2
GðsÞ ¼ ¼ 2 (9.6.4)
UðsÞ ms þ k
mω2
and finding the corresponding Gð jωÞ ¼ , which is real, its modulus therefore
mω2 þ k
equals the function itself. Then, from Eq. (9.6.4) and Eq. (7.3.5) the magnitude of the steady-
state response of the given system under a sinusoidal input is

mω2
X ¼ UjGð jωÞj ¼ UGð jωÞ ¼ U (9.6.5)
mω2 þ k

In order to derive the expression for X, we assume that the supporting structure is
stationary, but there is a virtual force f, applied at the free end of the microcantilever
beam, that causes the same out-of-plane bending as did the sinusoidal motion of the
supporting structure. As derived from literature on the strength of materials, a force applied
to the free end of a cantilever beam produces a deflection:
728 9 Application Problems

fl3
x¼ (9.6.6)
3EI

where f is the applied force, l is the length of the beam, E is elasticity modulus (Young’s
modulus), and I is the moment of inertia. EI is often called the bending stiffness.
wd3
For the given microcantilever beam, I ¼ , where w and d are the width and thickness of
12
the cantilever beam (see Figure 9.6.2(a)); hence, Eq. (9.6.6) becomes

f L3 4f L3
x¼  3 ¼ (9.6.7)
wd Ewd3
3E
12

Using Eq. (9.6.7) to express f in the frequency domain and substituting the value of X as
derived in Eq. (9.6.5), we obtain

Ewd3 Ewd3 mω2

F¼ X ¼ U (9.6.8)
4L3 4L3 ðmω2 þ kÞ

The microcantilever beam is experiencing a deformation and, consequently, strain. This

strain exists on both faces of the beam. Assuming that face 1 extends under bending while
face 2 compresses, their strains will have opposite signs – positive for face 1, and negative for
face 2. Working under the previous assumption of a virtual force f applied at the free end of
the microcantilever beam, we find the bending moment caused by this force Mb ¼ f L.
According to Hooke’s law, the stress due to bending, perpendicular to any cross-section
wd of the microcantilever beam, is related to the axial strain in the following manner:

σ ¼ Eε (9.6.9)

where σ is the normal stress, E is Young’s modulus, and ε is the axial (aligned with the
generating force) strain.
The stress is computed using Navier’s equation:

Mb d
σ¼ (9.6.10)
2I

where Mb is the bending moment due to the assumed virtual force f, d is the thickness of the
beam, and I is the cross-sectional moment of inertia about the bending axis.
For a rectangular cross-section beam of width w and thickness d, this moment of inertia is

wd3
I¼ (9.6.11)
12
 9.6 Sensors, Electroacoustic, and Piezoelectric Devices 729

Then, the magnitude of strain from Eqs. (9.6.8)–(9.6.10) is derived as

σ Mb d=2I f Ld f Ld 6f L
ε¼ ¼ ¼ ¼ ¼ (9.6.12)
E E 2IE 2ðwd =12ÞE Ewd2
3

Converting Eq. (9.6.12) into the frequency domain and substituting the expression for the
virtual force from Eq. (9.6.8) into it, we obtain

6L εwd3 mω2 3dmω2
ε¼ U ¼ U (9.6.13)
εwd2 4L3 ðmω2 þ kÞ 2L2 ðmω2 þ kÞ

Equation (9.6.13) relates the strain to be sensed to the frequency and magnitude of the
sinusoidal input. Now, we need to derive the expression relating this strain to the variation of
the voltage output vout of the Wheatstone bridge.
For a strain gauge connected to an electrical resistance, the following resistance–strain
relationship exists:
ΔR
¼ Kg ε (9.6.14)
R

where ε is the axial strain, Kg is the strain-gauge sensitivity, R is the electrical resistance, and
ΔR is the change in electrical resistance due to the strain.
Assuming that the four strain gauges attached to the microcantilever beam are identical,
the following relationship between the input voltage vin and the variation of the voltage
output vout exists (see Chapter 4 for derivation):

1 ΔR1 ΔR2 ΔR3 ΔR4
Δvout ¼  þ  vin (9.6.15)
4 R1 R2 R3 R4

From Eq. (9.6.14):

ΔR1 ΔR2 ΔR3 ΔR4

¼ ¼ ¼ ¼ Kg ε (9.6.16)
R1 R2 R3 R4

Therefore, Eq. (9.6.15) becomes

Δvout ¼ Kg εvin (9.6.17)

Converting to the frequency domain and using the expression for the strain described by Eq.
(9.6.13), the relationship between input parameters – the amplitude and frequency – and the
system parameters is derived:

3dmω2
ΔVout ¼ Kg UVin (9.6.18)
2L2 ðmω2 þ kÞ
730 9 Application Problems

Equation (9.6.18) constitutes the frequency domain model of the displacement sensor,
consisting of a microcantilever beam and four strain gauges connected in a Wheatstone bridge.
Recalling that the variation of the voltage output of the Wheatstone bridge is proportional
to the mechanical displacement being measured (Eq. (9.6.1)), we obtain the expression of
sensitivity as a function of the input frequency:

3dmKg ω2 Vin
K¼ (9.6.19)
2L2 ðmω2 þ kÞ

For a numerical simulation consider the following values of system parameters:

Microcantilever-beam geometry: L ¼ 200  106 m; d ¼ 100  109 m; w ¼ 40  106 m

ρ ¼ 5200 kg=m3

Electrical and mechanical parameters: Kg ¼ 2:2; E ¼ 1:6  1011 N=m2 ; vin ¼ 10 V

From the tables in Chapter 3 and the literature on the strength of materials, the equivalent
mass and equivalent stiffness of the microcantilever beam are found to be as follows:

33 33
meq ¼ m¼ Ldwρ ¼ 9:81  1013 kg
140 140

3EI 3E wd3
keq ¼ 3 ¼ 3 ¼ 2  104 N=m
L L 12

Only below-resonance input frequencies need to be considered. As derived in Chapter 7,

rffiffiffiffiffiffi
ke
the natural frequency of this second-order system is ωn ¼ ¼ 14281:6 rad=s
me
The numerical simulation results are presented in Figure 9.6.3, and the Mathematica code
for generating these results is straightforward, as shown in Figure 9.6.4.
As seen in Figure 9.6.3, the sensitivity of this displacement sensor increases with the
increase of input frequency and goes to infinity when the system approaches resonance.
Using this sensor for detecting sinusoidal motion with a close-to-resonant frequency is not
recommended since the very high sensitivity will cause a significant increase in the output
voltage of the Wheatstone bridge, thus damaging the strain gauges.
An accelerometer, as its name suggests, is used to measure the acceleration. This measure-
ment serves as an input for many types of control systems. There are different types of
accelerometers, depending on the physical effect employed. An electromechanical acceler-
ometer is shown in Figure 9.6.5, where A indicates an accelerometer device, and B denotes
the element that experiences the acceleration that needs to be measured. The accelerometer is
attached to the tested element.
The mechanical subsystem of this device is represented by a mass m (seismic or proof mass)
attached to the accelerometer’s body with a linear spring of stiffness k and a linear damper
 9.6 Sensors, Electroacoustic, and Piezoelectric Devices 731

Figure 9.6.3 Sensitivity of a displacement sensor vs. the sinusoidal input frequency

Figure 9.6.4 Sensitivity of a displacement sensor vs. the sinusoidal input frequency: Mathematica code
732 9 Application Problems

Figure 9.6.5 Electromechanical accelerometer

element with a damping coefficient c. The damping element may be represented by a viscous
fluid rather than a dashpot. The electrical system that is responsible for the actual measure-
ments is represented by a potentiometer P. In some implementations it may also include an
amplifier.
When the tested element B starts moving with acceleration X€ ðtÞ, it pulls the attached
accelerometer A and the proof mass along with it. If the displacement of the proof mass is
xðtÞ, and the displacement of the tested element is X ðtÞ, then applying the Newton’s second
law to the proof mass, we obtain
   
m€ _  X_ ðtÞ þ k xðtÞ  X ðtÞ ¼ 0
x ðtÞ þ c xðtÞ (9.6.19)

Let the relative displacement of the proof mass be xm ðtÞ ¼ xðtÞ  X ðtÞ. Then, Eq. (9.6.19)
becomes

mX€ ðtÞ þ m€
x m ðtÞ þ cx_ m ðtÞ þ kxm ðtÞ ¼ 0 (9.6.20)

Separating the components that contain the relative displacement at the left-hand side, we
obtain
x m ðtÞ þ cx_ m ðtÞ þ kxm ðtÞ ¼ mX€ ðtÞ
m€ (9.6.21)

If the potentiometer is calibrated to read this relative displacement, its output voltage is
linearly dependent on it as follows (see Section 4.2.3, Example 4.2.4 for detailed derivations):
xm
ν ¼ νin ¼ KL xm (9.6.22)
L
νin
where L is the potentiometer length, νin is the input voltage, and KL ¼ is the potentiometer
L
gain.
 9.6 Sensors, Electroacoustic, and Piezoelectric Devices 733

Figure 9.6.6 Electromechanical accelerometer: block diagram

Thus, the accelerometer dynamic system is modeled by two governing equations: Eqs.
(9.6.21) and (9.6.22). Its block diagram, considering that the system input is the acceleration
of the tested element X€ ðtÞ and the output is the voltage readout, is shown in Figure 9.6.6.
Depending on the selection of the system parameters m, c, and k, the device can be used as
an accelerometer as discussed above, or as a vibrometer that measures the amplitude of
sinusoidal displacement. The most commonly known application of a vibrometer is
a seismograph – the device to measure ground motion resulting from an earthquake.

Example 9.6.2 Modeling of a vibrometer and accelerometer

For the accelerometer discussed above, in order to measure the amplitude of sinusoidal displace-
ment the amplitude of the device’s response must be close to the amplitude of the displacement
input.
As derived in Chapter 7 (Eqs. (7.3.5) and (7.3.6)), the amplitude of a steady-state frequency
response of a second-order system equals r0 jGð jωÞj, where r0 is the input amplitude and
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
jGðjωÞj ¼ ½ReðGð jωÞÞ2 þ ½ImðGð jωÞÞ2 is the magnitude of the system transfer function in the
frequency domain.
For the electromechanical accelerometer discussed above, the transfer function can be written
as

KL s2
GðsÞ ¼ (9.6.23)
c k
s2 þ s þ
m m
c k
Recalling that ¼ 2ξωn and ¼ ωn 2 , this transfer function is converted to the frequency
m m
domain as follows:

KL ω2
GðjωÞ ¼ (9.6.24)
ðωn 2  ω2 Þ þ 2ξωn j

Multiplying the numerator and denominator by the complex conjugate of the denominator, and
separating the real and imaginary parts, we obtain
ωn 2  ω2 2ξωωn
GðjωÞ ¼ KL ω2  KL ω2 j (9.6.25)
ðωn 
2 ω2 Þ2 þ 4ξ 2
ω2 ωn 2 ðωn  ω2 Þ2 þ 4ξ 2 ω2 ωn 2
2
734 9 Application Problems

Then, using the expressions for the natural frequency and damping ratio, the magnitude of this
transfer function in terms of the system physical parameters m, k, and c is

KL ω2
jGðjωÞj ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (9.6.26)
ðk=m  ω2 Þ2 þ ðc=mÞ2 ω2

In order for the device to work as a vibrometer r0 jGðjωÞj ffi r0 , hence jGðjωÞj ffi 1 for the specific
frequency(ies) of interest. Otherwise, this sensor will be functioning as an accelerometer.
This can be illustrated by the following numerical simulation.
c k
Assume the following values of the system parameters: ¼ 115; ¼ 70; KL ¼ 2:15; ω ¼ 60.
m m
Then, jGðjωÞj ¼ 0:9986 ffi 1, which means that the magnitude of the system steady-state frequency
response r0 jGðjωÞj is almost equal to the magnitude r0 of the sinusoidal displacement input with
the frequency of 60 rad/s – the sensor works as a vibrometer.
c k
Changing the first two parameters as ¼ 345; ¼ 210, the magnitude jGðjωÞj decreases to
m m
0.369, making the sensor work as an accelerometer for the sinusoidal input of the same frequency.
Assuming that r0 ¼ 1, for simplicity, compute the acceleration of this sinusoidal input as 3600.
Then, the amplitude of response r0 jGðjωÞj ¼ 0:369 is 1:02  104 times the acceleration of the
input. Hence, this device works as an accelerometer with the gain 1:02  104 .

In addition to the described spring–damper construction, accelerometers can employ

other technologies that use the change of some physical property due to acceleration of the
studied element as a measure of the acceleration that is experienced by that element. For
example, in a capacitive device, the degree of acceleration is indicated by the change of
capacitance, while in a piezoelectric one it is measured by the change in the voltage output of
the piezoelectric crystal, and in a piezoresistive device the degree of acceleration is measured
by the change in resistance.
Capacitive accelerometers are typically constructed as microdevices or MEMS (micro-
electromechanical systems). The accuracy and high sensitivity of MEMS accelerometers, as
well as their very small size (micrometers in size in silicon) and low cost make such devices
very attractive to the industry. Currently, the biggest market demand for such devices is
exerted by the automotive industry, the manufacturers of personal electronic devices, and the
aerospace industry. MEMS accelerometers are used in numerous applications such as image
stabilization in camcorders and built-in cameras, some user-interface control in smartphones
and tablets, navigation units and anti-skidding systems in cars, contactless game controllers,
and real-time controls for smart weapon systems.
A capacitive accelerometer is illustrated in Figure 9.6.7. The assembly is used for
a balanced force micro-accelerometer (MEMS device), widely used in the air-bag deployment
sensor in automobiles. Since there is no room in a microdevice for a coil spring, damper, or
 9.6 Sensors, Electroacoustic, and Piezoelectric Devices 735

B
B

C
Beam motion
E1 A
A B E1 E2
E2 C
C

Acceleration

Figure 9.6.7 Balanced force micro-accelerometer: (a) the device configuration at rest; (b) the configur-
ation in response to an applied acceleration

even a seismic mass, miniaturizing an accelerometer to the microscale requires a different

approach. In the device shown in Figure 9.6.7 a plate beam A serves as a substitute for the
seismic mass. It is supported by the two folded tethers (denoted by B), which are firmly fixed
to the supporting plate. These tethers play the role of the spring, and the surrounding air
provides damping.
The capacitance C is measured between the pair of electrodes E1 and E2, where E1 is the
stationary electrode, and E2 is the moving one. When an acceleration is applied to the device,
the plate beam moves in the direction opposite to the acceleration, and the distance between
the electrodes E1 and E2 changes, thus changing the capacitance C. The relation between the
change in the measured capacitance and the experienced acceleration depends on the
geometry and materials of the accelerometer, and the device scale is calibrated accordingly.
Piezoelectric and piezoresistive accelerometers are presented in the next section.

9.6.2 Piezoelectric Devices

Piezoelectric crystals that are at the core of piezoelectric devices represent a subset of a wider
class of materials that possess an oriented internal structure – electrets. Electrets are usually
solid dielectrics that have a permanent, fixed electrical polarization, just like permanent
magnets have a magnetic polarization. The molecules of an electret are dipoles, aligned in
such a way that the material possesses opposing electrical poles of positive and negative
charges of the same magnitude. The existence of these poles results in a discernible fixed
electrostatic surface charge, or static potential. The real-charge electrets exhibit a relatively
strong external electric field, while dipolar-charge electrets are electrically neutral even
though they still possess aligned dipolar charges. The charge and dipole orientation of an
electret depend on many factors, such as, for example, the chemical constitution of the
material, macromolecular arrangements, the presence of impurities, etc.
While both electret types are widely used in different applications such as, for example,
electret microphones, headphones, speakers, transducers, air filters, and other engineering
devices, the real-charge ones have a disadvantage. Exposed to the air, this type of electret
736 9 Application Problems

attracts the charged particles floating in the air as ions and dust. The accumulation of these
particles neutralizes the surface charge of this electret, rendering it unusable.
A piezoelectric crystal is an electret in which the unit cells of its lattice contain electric
dipoles; these unit cells are oriented such that the dipoles are aligned. There is no net charge
on the faces of a piezoelectric crystal since the electric dipole moments exactly cancel out.
Upon the application of a sufficient force, a piezoelectric crystal deforms, causing
a disruption in the alignment of its unit cells, resulting in a re-orientation of the electric
dipoles. This misalignment forces the dipole moments out of balance, subsequently generat-
ing an electrical charge on the surface of the crystal. Therefore, a voltage is developed across
the piezoelectric crystal: mechanical energy due to the applied force is converted into
electrical energy.
The described piezoelectric effect is employed in sensors that measure strains, pressures,
and forces.
The inverse piezoelectric effect performs the opposite energy conversion: a voltage, applied
to a piezoelectric crystal, causes its deformation, thus producing a force. It is used primarily
in actuators. Among the everyday devices that use the inverse piezoelectric effect are quartz
watches. In these watches, the electrical energy, which is supplied by a battery, makes the
piezoelectric crystal oscillate with a very high frequency. Then, the electronic circuitry of the
watch modifies the frequency of these oscillations to one hertz (Hz, one cycle per second) and
uses them as an input to the gear system that drives the watch hands around the clock face.
Piezoelectric materials can be monocrystalline, such as, for example, quartz and zinc oxide,
or ferroelectric. Ferroelectric materials are represented by various ceramics (for example,
PZT – lead zirconate titanate – and barium titanate), and polymers. The ceramics, in
particularly PZT, are widely used in medical technology as ultrasonic transducers, which
convert electrical energy into high-frequency vibrations. Some of the best-known naturally
occurring piezoelectric crystals are quartz, Rochelle salt, and sucrose (cane sugar).
The piezoelectric effect of monocrystalline materials depends on their crystalline struc-
tures, and it is relatively small, while the effect for ferroelectrics is considerably more
pronounced, which makes them a better fit for the majority of applications. Additionally,
forcing an inclusion of metal ions (for example, nickel, bismuth, or niobium) into the
crystalline lattice of piezoelectric ceramics such as PZT allows optimization of the piezoelec-
tric parameters of the ceramic according to the desired specifications. Nonetheless, ferro-
electrics have some specific disadvantages. Unlike the monocrystalline materials, the
fabricated ferroelectrics need to be polarized with a strong electrical field to gain the
piezoelectric capacity. Consequently, if exposed to an electric field opposite to their polar-
ization, the ferroelectrics can become depolarized, losing their piezoelectric qualities.
Exposure to high temperatures (greater than a specific threshold that depends on the
material) also causes depolarization.
The large modulus of elasticity (Young’s modulus) of piezoelectric materials is responsible
for the linear relationship between the force and the voltage for a wide range of inputs. It is
 9.6 Sensors, Electroacoustic, and Piezoelectric Devices 737

also responsible for the small deformation of such materials, which makes them particularly
well suited for micro- and nano-devices, but unfit for measuring static forces.
The moderate-to-low cost of piezoelectric devices, their small size, sufficient accuracy, and
the lack of need for an external power source make them very attractive for a wide spectrum
of applications, from simple applications such as spark lighters and inkjet printers to
sophisticated medical technology and scanning tunneling microscopes.
Piezoelectric accelerometers are typically used in civil structures that operate in dynamic
settings. The construction of such an accelerometer is shown in Figure 9.6.8(a), where
A denotes an accelerometer device and B indicates the element that experiences the acceler-
ation to be measured.
The accelerometer body is secured to the tested element. The piezoelectric crystal that
constitutes a core of this accelerometer is mounted to the accelerometer body and sand-
wiched between two metal plates (electrodes). These electrodes are connected to a voltage-
measuring device such as a voltmeter or potentiometer. The acceleration applied to B results
in the motion of the seismic mass m, which causes a deformation of the piezoelectric crystal.
The piezoelectric effect is responsible for the voltage v being generated across the crystal.
Within the operating envelope of the specific accelerometer the generated voltage is linearly
dependent on the experienced acceleration.
A piezoresistive accelerometer is illustrated in Figure 9.6.8(b), where A denotes an accel-
erometer device, B indicates the element that experiences the acceleration to be measured,
C is a silicon cantilever beam, and P denotes a piezoresistor. Like the balanced force micro-
accelerometer discussed earlier, the cantilever-beam accelerometer, shown in Figure 9.6.8(b),
is a MEMS device. Its operation is based on the piezoresistive effect – the resistivity of
a material changes in response to the applied mechanical stress.
In the design shown, the accelerometer body is secured to the tested element. When an
acceleration is applied to the tested element, the proof mass m starts moving, causing

Wheatstone bridge

A
v
m
A
P
Piezo- Voltmeter
electric v or C m
crystal Potentiometer
Constraint base
B B

(a) (b)

Figure 9.6.8 (a) Piezoelectric accelerometer; (b) piezoresistive accelerometer

738 9 Application Problems

a mechanical strain in the cantilever beam C. The deformation of the beam results in
a strained piezoresistor P and a consequent change in its resistivity. This change is detected
and measured by the electrical subsystem, typically represented by a Wheatstone-bridge
circuit.
The cantilever-beam accelerometer’s volume is often filled with a viscous fluid for damping
the motion of the proof mass.
The dependency between the strain, experienced by P, and the change of resistivity is
linear.
Piezoresistive accelerometers are widely used for low-frequency vibrations since they can
measure very small accelerations. They also show satisfactory performance in shock envir-
onments. These devices exhibit high sensitivity and typically do not require pre-
amplification. The major disadvantage of piezoresistive accelerometers is temperature sensi-
tivity and limited high-frequency response.

Example 9.6.3 A piezoelectric accelerometer

A piezoelectric accelerometer to measure an amplitude of sinusoidal displacement input is shown
in Figure 9.6.9(a). This representation assumes that the piezoelectric plate is modeled as a mass–
spring–damper system and that the total system losses are lumped into a known single viscous
damping coefficient ce . The other known system parameters are the seismic mass m, spring
stiffness k, density of the piezoelectric crystal ρpz , Young’s modulus of the piezoelectric crystal
Epz , piezoelectric voltage constant gpz , and the amplitude v of the voltage read by the piezoelectric
sensor. The piezoelectric crystal is a thin square plate, a side of which measures w and its height
measures h.
The model of this system is an expression that relates the magnitude of the vibrating-platform
acceleration and magnitude of the sensed voltage v.
z

Vibrating me
k c
platform

u = U sin(ω t)
m u = U sin(ω t) ke ce

Vibrating platform
Piezoelectric
crystal
(a) (b)

Figure 9.6.9 Piezoelectric accelerometer: (a) schematic; (b) lumped-parameter model

 9.6 Sensors, Electroacoustic, and Piezoelectric Devices 739

Since the seismic mass is in contact with the piezoelectric crystal at all times, and they move
together, the lumped-parameter model is simplified to a single equivalent mass. The influence of
the piezoelectric crystal on the motion of this equivalent mass is represented by a spring with
a stiffness coefficient kpz . Therefore, the lumped-parameter model can be further simplified to
contain a single equivalent spring exerting the same effort as a combination of the given spring
k and the spring kpz . Note that the piezoelectric crystal is undergoing axial longitudinal
deformation.
As derived in the literature on the strength of materials and shown in Chapter 3, the equivalent
mass and equivalent stiffness, indicated in the lumped-parameter model shown in Figure 9.6.9(b),
are computed as follows:
8
> 1
<me ¼ m þ mpz ¼ m þ ρpz hw2
3 (9.6.27)
>
:ke ¼ k þ kpz ¼ k þ Epz w
2

where mpz and kpz are the equivalent mass and stiffness of the piezoelectric crystal respectively.
The governing equation for the mechanical subsystem of this dynamic system is derived using
the Newton’s second law:

_  ke ðz  uÞ
me z€ ¼ ce ð_z  uÞ (9.6.28)

where u is the oscillatory displacement of the vibrating platform, and z is the displacement of the
lumped mass me .
Denoting the relative motion of the lumped mass with respect to the vibrating platform as
x ¼ z  u, the governing equation is rewritten as follows:

d2
me ðx þ uÞ ¼ ce x_  ke x ) me x€ þ ce x_ þ ke x ¼ me u€ (9.6.29)
dt2

As stated in references on the dynamics of MEMS, the sensed voltage v is proportional to the
relative displacement x

v ¼ gpz Epz x (9.6.30)

Deriving x from Eq. (9.6.30) and substituting it into Eq. (9.6.29), we obtain

me ce ke
v€ þ v_ þ v ¼ me u€ (9.6.31)
gpz Epz gpz Epz gpz Epz

which generates the following transfer function:

VðsÞ gpz Epz me s2
GðsÞ ¼ ¼ (9.6.32)
UðsÞ me s2 þ ce s þ ke
740 9 Application Problems

One of the most important characteristics of an accelerometer is the transmissibility – the ratio
of the amplitude of the sensed voltage to the magnitude of displacement input. The transmissibil-
ity is a modulus of the transfer function GðjωÞ as expressed in the frequency domain. Recalling the
rffiffiffiffiffiffi
ke ce ce
relationship between the system and model parameters: ωn ¼ and ξ ¼ ¼ pffiffiffiffiffiffiffiffiffiffi (see
me 2me ωn 2 ke me
Chapter 7 for details), the expression for transmissibility is derived as follows:

gpz Epz ω2 gpz Epz ω2

T ¼ jGðjωÞj ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (9.6.33)
ðk=m  ω2 Þ2 þ ðc=mÞ2 ω2 ðωn 2  ω2 Þ2 þ 4ξ 2 ωn 2 ω2
ω
We divide numerator and denominator of jGðjωÞj by ωn 2 , and denote ratio r ¼ . Then, the
ωn
transmissibility becomes

gpz Epz r2
T ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (9.6.34)
ð1  r2 Þ2 þ ð2ξ rÞ2

The transmissibility, being the modulus of the transfer function GðjωÞ, can be viewed as the gain
of the Bode plot – a graph of the jGð jωÞj vs. frequency ω, where the frequency is in radians
per second, and jGð jωÞj is measured in decibels, jGðjωÞj; dB ¼ 20log10 ðjGðjωÞjÞ. The Bode plot
consists of two graphs: gain (jGð jωÞj vs. ω, where jGð jωÞj is related to the magnitude of the
frequency response) and phase (the phase shift of the frequency response). It is of utmost
importance for the system frequency response analysis and control design (see Chapter 7 for
more information).
Given the following numerical values of the system parameters:

Piezoelectric crystal parameters:

w ¼ 0:015 m; h ¼ 1:5  103 m; ρpz ¼ 7500 kg=m3

;
Epz ¼ 5  1010 N=m2 ; gpz ¼ 0:074 m3 =C; v ¼ 0:1 V

Seismic mass and supporting elements:

m ¼ 0:075 kg; k ¼ 80 N=m; ce ¼ 135 Ns=m;

the Bode plot is generated (gain only, see Figure 9.6.10).

The peak gain ðGmax Þ occurs at the resonant frequency ωr . The numerical values for the
resonant frequency and peak gain can be read from the graph or can be derived analytically
using the procedure presented in Section 7.3.4. The resonant frequency is computed as
 9.6 Sensors, Electroacoustic, and Piezoelectric Devices 741

Figure 9.6.10 Gain Bode plot of the piezoelectric accelerometer system

ωn
ωr ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi 3:14  105 rad=s; at the resonant frequency peak gain is Gmax ffi 236:3 dB. The
1  2ξ2
Mathematica code for constructing the Bode plot and analytically deriving the resonant frequency
and peak gain is shown in Figure 9.6.11.
An accelerometer is typically constructed to have a high natural frequency. Assuming that for
a given piezoelectric device ωn ≫ ω, the denominator of the expression for transmissibility
(Eq. (9.6.34)) becomes approximately equal to 1.
Since T ¼ jGðjωÞj, from Eq. (9.6.32) we can conclude that

jVðjωÞj
jGðjωÞj ¼ ¼T (9.6.35)
jUðjωÞj

Then, the relationship between the voltage reading and periodic input amplitude is

ω2
V ¼ TU ffi gpz Epz U (9.6.36)
ωn 2

where ω2 U ¼ A is the magnitude of acceleration of the sinusoidal displacement input.

Hence, the relationship between the sensed acceleration and the voltage reading becomes

ωn 2
A¼ V (9.6.37)
gpz Epz
ωn 2
where the constant is defined by the accelerometer construction.
gpz Epz
742 9 Application Problems

Figure 9.6.11 Generating a Bode plot and computing the resonant frequency and peak
gain: Mathematica code

Another important characteristic of an accelerometer is its very low damping. Computing the
damping ratio of the given piezoelectric accelerometer, we find ξ ¼ 0:00283. With such a low
damping ratio the resonant frequency is approximately equal to the natural frequency of the
system (which can be easily shown numerically), and is used to derive the numeric expression
relating A to V.
For a given device, Eq. (9.6.37) reads A = 26.73 V. If, as specified, the voltage reading equals
0.1 V, the acceleration of the measured sinusoidal input equals 2.673 m/s2.

9.6.3 Electroacoustic Devices

Electroacoustic devices are transducers that convert sound pressure into electrical signals
(microphones), or electrical signals into sounds waves (speakers).
 9.6 Sensors, Electroacoustic, and Piezoelectric Devices 743

Sound information may be visualized as a pattern of air pressure. Microphones convert

this information into a corresponding voltage pattern, while loudspeakers perform the
opposite operation. A concept of the fidelity of audio equipment describes the accuracy of
such transformations. High-fidelity audio systems generate minimal amounts of noise and
sound distortion, providing a high quality recording and reproduction of sound.
There are two types of microphones – passive and active. Passive microphones derive the
output solely from the absorbed acoustic power, while active microphones use an additional
external source of power to generate an output. Any passive microphone can be inverted to
make a speaker, although design modifications to adjust its acoustical and electrical charac-
teristics are often necessary. An active microphone typically has no speaker analogy.
Similarly, there are certain speaker designs that cannot be converted into a microphone.
A detailed discussion of technological approaches used for microphone and speaker
design belongs to the field of acoustic engineering and is beyond the scope of this book.
The subjects of interest in the current section are dynamic microphones and their analogs –
loudspeakers – and two types of condenser microphones.
A dynamic microphone, illustrated in Figure 9.6.12, is an example of a passive device and
one of the most commonly used acoustic-to-electric transducers. Its operation is based on the
phenomenon of electromagnetic induction, described earlier in Chapter 6.
As shown in Figure 9.6.12, a dynamic microphone has three major parts – coupled
magnets, a thin metallic diaphragm in the form of a cone, and a coil of wire (in some
literature sources it is called a voice coil) fastened to the diaphragm. The sound waves
represented by the changing air pressure cause movement of the cone, which in turn provides

Diaphragm (cone)

Microphone
S enclosure

Sound N
Magnets
waves
input N

S
Coil

Signal output
t
(voltage or current)

Figure 9.6.12 Dynamic microphone: schematic

744 9 Application Problems

for the motion of the coil in the magnetic field, generated by the two coupled permanent
magnets. According to Faraday’s law of induction, this change in the magnetic environment
of the coil (conductor) causes a voltage to be induced in the coil. The induced voltage is
proportional to the negative rate of change of the magnetic flux (see Section 6.3.1).
Modeling of this dynamic system starts with considering its mechanical and electrical
subsystems separately. The mechanical subsystem is represented by the diaphragm and the
coil. It can be approximated by a typical mass supported with a spring–damper assembly, as
shown in Figure 9.6.13(a), where m represents the combined mass of the diaphragm and the
coil, and k and c denote the stiffness and damping coefficients of the diaphragm, respectively.
These constants depend on the material and geometry of the diaphragm. The parameter
FS ðtÞ is the force exerted on the diaphragm by the incoming sound waves, and fM ðtÞ is the
force exerted on the coil by the magnetic field.
The electrical subsystem is represented by the same diaphragm and coil assembly. It is
modeled as a simple circuit, shown in Figure 9.6.13(b), where R is the combined resistance of
the diaphragm and the coil, L denotes the inductance of the coil solenoid, νb is the induced
voltage, and i ðtÞ is the coil current, which is also the current in the circuit.
We let the displacement of the diaphragm-coil assembly be xðtÞ. Then, applying the
Newton’s second law to the mechanical subsystem, we obtain

m€ _ þ kxðtÞ ¼ FS ðtÞ þ fM ðtÞ

x ðtÞ þ cxðtÞ (9.6.38)

Recalling the discussion on magnetic forces in Section 6.3.1, we derive the expression for
fM ðtÞ using Eq. (6.3.9):

fM ðtÞ ¼ nlBiðtÞ (9.6.39)

where B is the magnetic flux density, iðtÞ is the induced current, l is the length of a single coil
segment, and n denotes the number of loops in the coil.
For the depicted microphone it can be safely assumed that the magnetic flux density is
uniform, and perpendicular to the current vector.

Figure 9.6.13 Dynamic microphone: (a) mechanical and (b) electrical subsystems
 9.6 Sensors, Electroacoustic, and Piezoelectric Devices 745

Since B, l, and n are constants, KM ¼ nlB. Then, the governing equation for the mechanical
subsystem becomes

m€ _ þ kxðtÞ ¼ FS ðtÞ þ KM iðtÞ

x ðtÞ þ cxðtÞ (9.6.40)

As derived earlier (Eq. (6.3.13)), the induced voltage is proportional to the velocity of the
conductor, moving in the magnetic field:

_
vb ¼ Kb xðtÞ (9.6.41)

where Kb is a constant coefficient that depends on the coil geometry and magnetic flux
density.
Since there is no voltage input in the system, the governing equation for the electrical
subsystem is

di
Ri þ L þ vb ¼ 0 (9.6.42)
dt

The microphone dynamics is modeled by Eqs. (9.6.40)–(9.6.42).

If the input into the system is the force exerted by the sound waves, and the output is the
induced voltage, then the system block diagram and transfer function representations are
derived as follows.
Taking the Laplace transform of the governing equations, we obtain
8 
< ms þ cs þ k X ðsÞ ¼ FS ðsÞ þ KM IðsÞ
> 2

Vb ¼ Kb sX ðsÞ (9.6.43)
>
:
ðR þ LsÞIðsÞ þ Vb ¼ 0

The expression for Vb is substituted into the last line of Eq. (9.6.43) and we solve for
IðsÞ, expressing it in terms of X ðsÞ. Then, we substitute the result into the first line of Eq.
X ðsÞ
(9.6.43) and derive the transfer function . The resulting block diagram is shown in
FS ðsÞ
Figure 9.6.14.
The derivation of the system transfer function from this block diagram is straightforward:

Fs (s) R + Ls X(s) Vb (s)

Kb s
2
(R + Ls)(ms + cs + k) + KM Kb s

Figure 9.6.14 Dynamic microphone with voltage output: block diagram

746 9 Application Problems

Fs (s) R + Ls X(s) Kb s I(s)

−
(R + Ls)(ms 2 + cs + k) + KM Kb s R + Ls

Figure 9.6.15 Dynamic microphone with current output: block diagram

Figure 9.6.16 Loudspeaker: (a) mechanical and (b) electrical subsystems

Vb ðsÞ ðR þ LsÞKb s
¼ (9.6.44)
FS ðsÞ ðR þ LsÞðms2 þ cs þ kÞ þ KM Kb s

If the system output is the induced current in the coil, the block diagram is as shown in
Figure 9.6.15, and the corresponding system transfer function is

IðsÞ Kb s
¼ (9.6.45)
FS ðsÞ ðR þ LsÞðms þ cs þ kÞ þ KM Kb s
2

Since the induced voltage is proportional to the speed of motion of the diaphragm–coil
assembly, dynamic microphones are also known as velocity-sensitive microphones or simply
as velocity microphones. They are low-noise microphones, require no external power inputs,
have a relatively rugged construction, and can be miniaturized easily. Combined with
a reasonably low cost, these advantages make dynamic microphones very attractive for
broadcasting and sound recording. The disadvantage of dynamic microphones is their low
sensitivity, which calls for an obligatory signal amplification.
The inverse of a dynamic microphone is a loudspeaker. The speaker construction is the
same as depicted in Figure 9.6.12, only the input is an electric signal, and the output is the
sound waves. The dynamic system of a loudspeaker is shown in Figure 9.6.16.
The only force acting on the diaphragm–coil assembly is the magnetic force fM ðtÞ.
Similarly to dynamic microphones, the expression for fM ðtÞ is derived using Eq. (6.3.9):

fM ðtÞ ¼ nlBiðtÞ (9.6.46)

 9.6 Sensors, Electroacoustic, and Piezoelectric Devices 747

where B is the magnetic flux density, iðtÞ is the circuit current, l is the length of a single coil
segment, and n denotes the number of loops in the coil.
As before, letting KM ¼ nlB, we obtain the governing equation for the mechanical
subsystem:

m€ _ þ kxðtÞ ¼ fM ðtÞ ¼ KM iðtÞ

x ðtÞ þ cxðtÞ (9.6.47)

The governing equation for the electrical subsystem is

di
Ri þ L þ vb ¼ vðtÞ (9.6.48)
dt

Considering the expression derived earlier for the induced voltage (Eq. (9.6.41)), the
loudspeaker model is a set of three equations: Eqs. (9.6.47), (9.6.48), and (9.6.41). In the
Laplace domain it becomes
8 2 
< ms þ cs þ k X ðsÞ ¼ KM IðsÞ
>
Vb ¼ Kb s X ðsÞ (9.6.49)
>
:
ðR þ LsÞIðsÞ þ Vb ¼ VðsÞ

Using the same procedure as described earlier for the dynamic microphone, the block
diagram representation is constructed and the system transfer function derived. For the
speaker system there is a single option of the input–output pair – the applied voltage is the
input, and the generated sound signal, represented by the displacement of the diaphragm, is
the output.
The block diagram is shown in Figure 9.6.17, and the corresponding system transfer
function is

X ðsÞ KM
GðsÞ ¼ ¼ (9.6.50)
VðsÞ ðR þ L sÞðms2 þ cs þ kÞ þ KM Kb s

Even though this speaker is an inverted dynamic microphone, its proportions and physical
construction are dissimilar, since this device is optimized for the different set of electrical and
acoustical characteristics.

V(s) ms 2 + cs+k I(s) KM X(s)

(R + Ls)(ms 2 + cs + k) + KMKb s ms 2 + cs + k

Figure 9.6.17 Loudspeaker: block diagram

748 9 Application Problems

Example 9.6.4 Dynamic response of a loudspeaker

The derived loudspeaker transfer function presented in Eq. (9.6.50) is of third order. Given some
numerical values for the system parameters and voltage input:

Figure 9.6.18 Dynamic response of the loudspeaker: (a) Mathematica code; (b) response graph
 9.6 Sensors, Electroacoustic, and Piezoelectric Devices 749

KM ¼ 16:0 N=A; R ¼ 12:1 Ω; L ¼ 103 H; Kb ¼ 12:5 Vs=m

m ¼ 0:002 kg; k ¼ 5:9  105 N=m; c ¼ 0:02 Ns=m; v ¼ 0:5 V

We can derive the dynamic response using the Mathematica functions TransferFunctionModel
and OutputResponse.
The code for generating the dynamic response is shown in Figure 9.6.18(a) and the response
itself is shown in Figure 9.6.18(b).

As seen from the graph in Figure 9.6.18(b), this system is fast – it reaches the steady state in
approximately 6 milliseconds.
The input voltage to a loudspeaker can be viewed as an electrical “image” of the sound,
emitted by an audio signal source, typically either a recording or a microphone. This
electrical image is a signal that has the same frequency, amplitude, and waveform (harmonic
content) as the original sound. The constant voltage input used for this simulation is just
a single component of that image. When a voltage is applied to the voice coil of the
loudspeaker, the electrical current starts flowing through this coil in the magnetic field.
A magnetic force is generated and the voice coil starts moving, driving the cone of the
loudspeaker and producing sound in the air. One of the conditions of reproducing the sound
pressure variations of the original signal is a fast convergence of the system response. To
better visualize the importance of fast convergence, a simulation demonstrates the system
response to a piecewise-continuous input that is a square-wave approximation of a small
segment of a sound wave (see Figure 9.6.19).
As the simulation results show, while a response convergence to the input is, in
general, achieved, significant imperfections are observed. They are particularly pro-
nounced at the points of input discontinuity. Also, when the duration of some constant
input is small (for example, inputs applied from 0 to 0.01 s. and from 0.03 s to 0.04 s),
the response still exhibits visible oscillations instead of converging to a constant steady
state.
Returning to the transfer function of this loudspeaker system, we compute its poles using
the Mathematica function TransferFunctionPoles obtaining the following results:
p1 ¼ 9617:4; p2;3 ¼ 1246:28 19225j. This dynamic system has one real pole and two
complex ones that are complex conjugates. Since the real parts of all three poles are strictly
negative, the system is stable (see Section 7.2.2 for details). The ratio of the real parts of these
Reðp1 Þ  9617:4
poles ¼ ≃ 7:7 shows that the complex poles are dominant, and this
Reðp2;3 Þ  1246:28
system can be approximated by the second-order system. The transfer function of such
approximation is derived in the following way (see Section 7.1.6 for details).
750 9 Application Problems

Figure 9.6.19 Dynamic response of the loudspeaker to a piecewise-continuous input

For a third-order transfer function

aωn 2
GðsÞ ¼
ðs  aÞðs2 þ 2ξωn s þ ωn 2 Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffi
where the complex conjugate poles pcc ¼ ξωn  jωn 1  ξ 2 are dominant, the second-
order system approximation transfer function is

ωn 2
Gappr ðsÞ ¼
s2 þ 2ξωn s þ ωn 2

(the contribution of the nondominant real pole is neglected, and the numerator of the
transfer function is divided by the magnitude of the nondominant pole).
Therefore, for the original transfer function of the given loudspeaker dynamic system:

8  106 8  106
GðsÞ ¼ ¼
s3 þ 12110s2 þ 3:95  108 s þ 3:57  1012 ðs þ 9617:43Þðs2 þ 2492:57s þ 3:71  108 Þ

the approximation is
8  106 =9617:43 831:823
Gappr ðsÞ ¼ ¼
s2 þ 2492:57s þ 3:71  108 s2 þ 2492:57s þ 3:71  108

This approximation is possible, but since the ratio of the real parts of the dominant and
nondominant poles is less than 10, the quality of such an approximation is not very good,
which is evidenced by the simulation results in Figure 9.6.20.
 9.6 Sensors, Electroacoustic, and Piezoelectric Devices 751

Figure 9.6.20 Dynamic response of the loudspeaker: actual system vs. a second-order approximation

If the inductance of the given loudspeaker system was so small that it could be neglected,
the system would become second order instead of the third order, with the transfer function

KM 661:16
G2nd ðsÞ ¼ ¼
Rðms2 þ cs þ kÞ þ KM Kb s s2 þ 8274:5s þ 2:95  108

Its settling time decreases (see Figure 9.6.21(a)), and it will follow a piecewise-continuous
input much more closely (see Figure 9.6.21(b)), even though there are still significant
imperfections at the points of input discontinuity. These “spikes” are higher than they
were for the original third-order system because of the higher overshoot of the second-
order system (see Figure 9.6.21(a)).
An amplifier is an integral part of any high-fidelity speaker, since the electrical signal input
needs to be large enough to provide for the coil vibrations of sufficient amplitude to
reproduce the sound pressure pattern of the recorded signal.
Some important considerations of a good speaker design include an even frequency
response so that no frequency is unduly weakened or emphasized, and a minimal sound
distortion, i.e., minimal creation of new frequencies in the output signal. To achieve the
typical frequency range of the high-fidelity reproduction of sound (50 Hz to 15 kHz) in
a single speaker is difficult since a flawless generation of high and low frequencies requires
different designs of the diaphragm. To produce high frequencies, a loudspeaker diaphragm
needs to be very light to be able to rapidly respond to the input signal. For low frequencies,
a large and relatively heavy diaphragm performs best. This calls for more power to be
752 9 Application Problems

Figure 9.6.21 Dynamic response of the loudspeaker: (a) original system vs. the second-order system; (b)
response of the second-order system to a piecewise-continuous input
 9.6 Sensors, Electroacoustic, and Piezoelectric Devices 753

MICROPHONE SPEAKER
Signal input

Sound Sound
Input
waves waves
input Common output
Output

AMPLIFIER

Listen
Listen

Talk Talk
Talk−Listen Talk−Listen
switch Signal output switch

Figure 9.6.22 Simple intercom assembly

supplied to make this massive structure vibrate. Additionally, since human sound perception
discriminates against lower frequencies, they require more acoustic power to be sufficiently
audible. Hence, the customary design approach involves subdividing the full range of sound
into two or three overlapping parts, and using multiple speakers, each optimized for its own
frequency range. An assembly, which is relatively inexpensive but of sufficient sound quality,
has two speakers, one covering the 50–1000 Hz range, the other the 500–15 kHz range. The
higher-end speakers employ at least three-speaker assembly – a low-frequency (woofer),
high-frequency (tweeter), and mid-range speaker to provide for a smoother frequency
response.
In a simple intercom assembly, a small dynamic speaker, described above, can be used at
both ends – as a microphone and as a loudspeaker, as illustrated in Figure 9.6.22.
Depending on the position of the “talk–listen” switch at the end device of the depicted
intercom, this device acts either as a microphone or a speaker. A signal from the microphone
end serves as an input for the amplifier (typically an op-amp, as described in Section 4.7), and
the resulting amplified electrical signal becomes an input into the speaker end of the
intercom. The low cost, relatively rugged construction, and simplicity of this intercom
assembly make it very attractive for a variety of uses, despite its at-best average sound
quality.
A condenser microphone, illustrated in Figure 9.6.23, is an example of an active device. Its
superior frequency response makes it the microphone of choice for many sound-recording
applications despite its high cost and the need for an external power supply.
754 9 Application Problems

Signal output

R Backplate
v(t)

+ + C R
Battery
vb(t)
Sound −
waves
input Thin metallic i(t)
diaphragm
−

Insulator

(a) (b)

Electret diaphragm

Backplate Pre-amplifier
(transistor)
Air cavity
Sound
waves
input
Signal output

Insulator

(c)

Figure 9.6.23 Condenser microphone: (a) traditional design – schematic; (b) traditional design – electrical
subsystem; (c) electret-based design

In a condenser microphone the sound pressure changes a distance between two charged
plates, subsequently changing the capacitance. This approach to microphone design employs
two major approaches: (a) traditional, where the capacitor is represented by a diaphragm
and a metal backplate (see Figure 9.6.23(a)), and (b) electret-based (see Figure 9.6.23(c)).
In a traditional condenser microphone, the diaphragm and the backplate form
a capacitor, which is charged by a battery as shown in Figure 9.6.23(a). As discussed in
Section 4.1.2, the capacitance of a parallel-plate capacitor depends on its geometry and the
material, and is represented mathematically with Eq. (4.1.13). The sound pressure, acting on
the microphone, moves the diaphragm, modifying the distance between it and the backplate.
 9.6 Sensors, Electroacoustic, and Piezoelectric Devices 755

Considering that, in a typical construction of this microphone, air is the dielectric substance
between the capacitor plates, the capacitance is expressed as

1 Q
C ¼ εA ¼ (9.6.51)
d vb

where Q is the stored charge, vb denotes the voltage provided by the battery, ε is the
permittivity, A is the area of the plates, and d is the distance between plates.
Since A, ε, and vb are constant, it can be stated that the charge on the capacitor plates is
inversely proportional to the distance between them:

1
Q ¼ εA vb (9.6.52)
d

When the distance between the diaphragm and the backplate changes, the stored charge
also changes, causing the electric current in the microphone circuit, shown in
Figure 9.6.19(b). From the definition of current (Eq. (4.1.1)):

ΔQ Δð1=dÞ
i¼ ¼ εAvb (9.6.53)
Δt Δt

Then, the voltage (output signal) measured across the resistor R (Eq. (4.1.9)) is

Δð1=dÞ
vðtÞ ¼ iR ¼ εAvb R (9.6.54)
Δt

and is proportional to the inverse of the distance between the diaphragm and the backplate
of the microphone. Hence, v(t) can be viewed as an image of the pattern of the sound pressure
that moves the diaphragm.
If we assign the following numerical values to the microphone parameters:
ε ¼ 8:854  1012 F=m; A ¼ 1  104 m2 ; R ¼ 100 MΩ; vb ¼ 0:2 V and assume that
the inverse of the distance between the diaphragm and the backplate changes in a periodic
fashion is sinðtÞ þ sinð0:5tÞ þ sinð2:5tÞ, the resulting image is shown in Figure 9.6.24.
Electret-based condenser microphones, as shown in Figure 9.6.23(c), use an electret
material for their diaphragms. Being very thin (with a thickness of the order of
a thousandth of an inch), an electret diaphragm is typically backed by a thin metallic
membrane.
Since electrets have a permanent fixed electrical polarization (see Section 9.6.2), there is no
need for an internal battery to charge the capacitor plates as in the traditional condenser
microphones – the backplate is charged by the electrical field of the electret diaphragm.
An electret-based microphone operation is the same as discussed above: the sound
pressure moves the diaphragm, thus changing the capacitance and causing an electric current
in the microphone circuit. The recorded output signal is an image of the applied sound. Since
756 9 Application Problems

Figure 9.6.24 Dynamic response of a condenser microphone

the capacitance of an electret-based microphone is small, the output signal requires amplifi-
cation. Hence, a pre-amplifier is an integral part of this microphone’s electrical subsystem.
While the recorded sound quality of electret-based microphones is average, these devices
are attractive due to their low cost, robust structure, and no need for an external power
source. Since electret materials can be dependably manufactured according to exact specifi-
cations, electret-based microphones can be custom-made for a specific application or pro-
ject. They also can be easily miniaturized, which prompts their main use for portable tape-
recorders, hearing aids, and other small-scale devices.
The major disadvantage of electret-based microphones lies in the fact that electrets may
lose their charge with time, thus making the device nonoperational.

REFERENCES
1. F. P. Beer, E. R. Johnston, Jr., P. Cornwell, and B. Self, Vector Mechanics for Engineers:
Dynamics, 11th ed., McGraw-Hill Education, 2015.
2. F. P. Beer, E. R. Johnston Jr., and D. F. Mazurek, Vector Mechanics for Engineers: Statics, 11th
ed., McGraw-Hill Publications, 2015.
3. J. L. Meriam, L. G. Kraige, and J. N. Bolton, Engineering Mechanics: Dynamics, 8th ed., Wiley,
2015.
4. Y. Cengel, M. Boles, and M. Kanoglu, Thermodynamics: An Engineering Approach, 9th ed.,
McGraw-Hill Education, 2018.
5. R. C. Hibbeler, Fluid Mechanics, 2nd ed., Pearson, 2017.
 References 757

6. N. Lobontiu and E. Garcia, Mechanics of Microelectromechanical Systems, Kluwer, 2004.

7. N. Lobontiu, System Dynamics for Engineering Students: Concepts and Applications, 2nd ed.,
Academic Press, 2017.
8. N. Lobontiu, Dynamics of Microelectromechanical Systems, Springer, 2007.
9. R. S. Figliola and D. E. Beasley, Theory and Design for Mechanical Measurements, 5th ed., John
Wiley & Sons, 2007.
10. W. J. Palm III, System Dynamics, 3rd ed., McGraw-Hill, 2014.
11. C. M. Close, D. K. Frederick, and J. C. Newell, Modeling and Analysis of Dynamic Systems,
Wiley, 3rd ed., 2002.
12. D. G. Luenberger, Introduction to Dynamic Systems: Theory, Models, and Applications, Wiley,
1979.
13. R. S. Esfandiari and B. Lu, Modeling and Analysis of Dynamic Systems, CRC Press, 2010.
14. R. L. Woods and K. L. Lawrence, Modeling and Simulation of Dynamic Systems, Prentice Hall,
1997.
 APPENDIX A
Units and Conversion Table

This appendix consists of five tables. Table A1 is about six base units of the International
System (SI) that are used in this book. Tables A2 to A4 list some derived SI units for
mechanical, electrical, magnetic, and thermal and fluid systems. Table A5 gives factors for
unit conversion between the SI and the US customary systems.

Table A1 SI base units*

Quantity Unit name Symbol

Length Meter m
Mass Kilogram kg
Amount of substance Mole mol
Time Second s
Temperature Kelvin K
Electric current Ampere A

* The International System of Units comprises a total of seven

base units. Besides the six units listed in Table A1, the base unit
that is not used in this text is the candela (cd) for luminous
intensity.

Table A2 Derived SI units for mechanical systems

Quantity Unit name Symbol Derivation

Velocity (v) Meter per second m/s 1 km/h = 0.2778 m/s

Kilometer per hour km/h
Acceleration (a) m/s2
Angle (θ) Radian rad 1 rad = 57.2958 deg
Degree deg 1 deg = 0.017453 rad
Angular velocity (ω) Radian per second rad/s 1 rad/s = 9.5493 RPM
Revolutions per minute RPM 1 RPM = 0.1047 rad/s
Agular acceleration (α) rad/s2
Mass moment of inertia (I) kg·m2
Mass density (ρ) kg/m3
Force (f) Newton N 1 N = 1 kg·m/s2
Moment of force (M) Newton-meter N·m
Torque (τ)

758
 Appendix A 759

Table A2 (cont.)

Quantity Unit name Symbol Derivation

ð
Impulse (I) Newton-second N·s I ¼ Fdt

Pressure (p, P) Pascal Pa 1 Pa = 1 N/m2

Mechanical stress (σ)
Energy Joule J 1 J = 1 N·m
Work
Power Watt W 1 W = 1 J/s
Coefficient of a translational N/m
spring (k)
Coefficient of a viscous N·s/m
damper (c)
Frequency (f) Hertz Hz 1 Hz = 1/s
Circular frequency (ω) rad/s

Table A3 Derived SI units for electrical and magnetic systems

Quantity Unit name Symbol Derivation

Electrical charge (q) Coulomb C 1 C = 1 A·s

Voltage (v) Volt V 1 V = 1 W/A
Electric potential (e)
Electrical resistance (R) Ohm Ω 1 Ω= 1 V/A
Electrical capacitance (C) Farad F
Electrical inductance (L) Henry H 1 H = 1 V·s/A
Electrical power Watt W 1 W = J/s
Magnetic flux Weber Wb 1 Wb = 1 V·s
Magnetic flux density Tesla T 1 T = 1 Wb/m2

Table A4 Derived SI units for thermal and fluid systems

Quantity Unit name Symbol Derivation

Temperature Celsius °C °C = K – 273.15

Heat (E) Joule J 1 J = 1 N·m
Heat flow rate ðqh Þ Watt W 1 W = 1 J/s
Thermal conductivity (k) W/(m·K)
W/(m·°C)
Convective coefficient (h) W/(m2·K)
W/(m2·°C)
Specific heat (c, cP , cV ) J/(kg∙K)
J/(kg∙°C)
760 Appendix A

Table A4 (cont.)

Quantity Unit name Symbol Derivation

Thermal capacitance (C) J/K
Thermal resistance (R) K·s/J
Volume (V) m3
Mass density (ρ) kg/m3
Volume flow rate (q) m3/s
Mass flow rate (qm) kg/s qm = ρq
Pressure (p, P) Pa 1 Pa = 1 N/m2
Liquid head (h) m
 
Specific gas constant Rg J/(kg·K)
Fluid capacitance (C) kg∙m2/N
Fluid resistance (R) Pa∙s/kg
Fluid inductance (I) kg/m4
Pneumatic capacitance (C) kg∙m2/N
Pneumatic resistance (R) Pa∙s/kg

Table A5 Unit conversion factors

Quantity SI unit to US unit US unit to SI unit

Length 1 m = 3.281 ft 1 ft = 0.3048 m

1 km = 1000 m = 0.62137 mile 1 mile = 5280 ft = 1609.34 m
Volume 1 liter (L) = 1 dm3 = 1.0564 quart 1 gal = 4 quarts = 3.875 L
Mass 1 kg = 1000 g = 0.06852 slug 1 slug = 14.594 kg
1 kg = 2.20462 lbm 1 lbm = 16 oz = 0.45359 kg
Force 1 N = 0.2248 lbf 1 lbf = 4.4484 N
Pressure 1 Pa ¼ N=m2 ¼ 0:9869  105 atm 1 atm = 101325 Pa
Energy 1 J = 1 N·m = 0.239 cal 1 cal = 4.184 J
1 J = 0.7376 lbf·ft 1 lbf·ft = 1.3558 J
Power 1 W = 1 J/s = 3.413 Btu/h 1 Btu/h = 0.293071 W
1 kW = 1000 W = 1.34102 hp 1 hp = 745.7 W
Temperature K = °C + 273.16 °F = 1.8 °C +32
°C = (°F − 32)/1.8 °F = 1.8 K – 459.69
Heat 1 kJ = 1000 J = 0.947817 Btu 1 Btu = 1055.056 J
 APPENDIX B
A Brief Introduction to MATLAB and Simulink

This appendix gives a brief introduction to MATLAB and Simulink. A detailed tutorial
about these software tools is available on the publisher’s website.
MATLAB is a programming platform designed specifically for engineers and scientists. It
provides an interactive environment for technical computation, data analysis, graphics, and
visualization. Simulink, which is built on top of MATLAB, is a block diagram environment
for modeling, simulation, analysis, and model-based design of multi-component multi-field
dynamic systems. It provides a graphical editor, customizable block libraries, and various
solvers for algebraic and differential equations.

B.1 MATLAB Operators and Functions

Listed in Tables B1–B9 are some commonly used MATLAB functions. In the Command
Window, typing help Func will display what the function named Func is, and how to use it.

B.2 Simulink Models

Because it is based on and integrated with MATLAB, Simulink is used with MATLAB. In
other words, to use Simulink, MATLAB must be started first. Once the MATLAB
Command Window is on, Simulink can be launched in one of the following two ways:

• Type simulink at the prompt; or

• Click on the Simulink icon on the MATLAB toolbar.

This will open a separate window – Simulink Start Page – from which one can create a new
Simulink model, work on an existing model, see some demo examples on modeling and
simulation with Simulink, and take other actions.

Model Window and Library Browser

To create a new Simulink model, click the “Blank Model” tab on the Simulink Start Page.
This opens a model window. On the toolbar of the model window, click the Library Browser

761
762 Appendix B

Table B1 Operators and special characters

+ Addition = Assignment operator

‒ Subtraction > Less than
* Multiplication <= Less than or equal to
/ Division > Greater than
\ Left division >= Greater than or equal to
^ Power == Equal to
.* Array multiplication operator ~= Not equal to
./ Array division operator & Logical operator: AND
.^ Array exponential operator | Logical operator: OR
% Percentage; comment ~ Logical operator: NOT

Table B2 Commonly used mathematical functions of MATLAB

Function Description Function Description

sqrt(x) Square root exp(x) Exponential

sin(x) Sine log(x) Natural logarithm
log10(x) Common logarithm
cos(x) Cosine abs(x) Absolute value
tan(x) Tangent sign(x) Signum function
cot(x) Cotangent max(x) Maximum value
asin(x) Arcsine min(x) Minimum value
acos(x) Arcosine real(x) Complex real part
atan(x) Arctangent imag(x) Complex imaginary part
sinh(x) Hyperbolic sine angle(x) Phase angle
cosh(x) Hyperbolic cosine conj(x) Complex conjugate
tang(x) Hyperbolic tangent

Table B3 Matrix functions

Function Utility

det Determinant
diag Diagonal matrices and diagonals of a matrix
eig Eigenvalues and eigenvectors
expm Matrix exponentials
eye Identity matrix
inv Matrix inverse
norm Matrix and vector norms
ones Matrix containing all elements ones
zeros Matrix containing all elements zeros
 Appendix B 763

Table B4 Functions for graphics

Function Utility

plot Generate xy plot

subplot Divide a graphic window into multiple ones
grid Add grid lines
legend Graph legend
title Place graph title
xlabel x-axis label
ylabel y-axis label

Table B5 Equation solvers

Function Description

linsolve Solve linear algebraic equations in matrix form

roots Find polynomial roots
fzero Find roots of nonlinear function of one variable
fsolve Solve system of nonlinear algebraic equations of multiple variables
ode45 Solve linear and nonlinear differential equations

Table B6 Transfer function formulation

Function Description

tf Create a transfer function model

zpk Create a zero-pole-gain model
series Series interconnection of the two blocks (transfer functions)
parallel Parallel interconnection of the two blocks (transfer functions)
feedback Obtain the transfer function of a feedback control system

Table B7 State-space formulation

Function Description

ss Create a state-space model

ss2tf State-space to transfer function conversion
tf2ss Transfer function to state-space conversion
764 Appendix B

Table B8 Time response

Function Description

impulse Impulse response

step Step response
lsim Response to arbitrary inputs

Table B9 Frequency response

Function Description

freqresp Frequency response

bode Bode plot of the frequency response
nyquist Nyquist plot
margin Gain and phase margins

Figure B2.1 A Simulink model window and the Library Browser

 Appendix B 765

Figure B2.2 The Commonly Used Blocks Library

icon to open another window showing a collection of Simulink block libraries; see Figure
B2.1. If a Simulink model has been created, the user can double-click the Simulink model file
that is stored in the computer. This will open an existing model window.
In Simulink, a model is to be constructed in the model window by using these Simulink
libraries. For instance, clicking on Commonly Used Blocks on the left panel opens the
window with a set of blocks as shown in Figure B2.2. As can be seen from the figure, the
blocks Gain, Sum, Integrator, and Product are similar to some components listed in
Table 6.5.1, but the library has many other blocks with different functionalities. Clicking
another entry on the panel, say Continuous, opens another library containing blocks like
Differentiator, Transfer Fcn, State-Space, and PID Controller. There are
many other block libraries on the panel.

Creation of a Model
As mentioned previously, a model is built in a model window. This is done by taking the
following actions:

1. Select a wanted block from a block library and place it in the model window. This is done
by first clicking and holding the block, and then dragging and dropping it anywhere in the
window.
766 Appendix B

2. Connect two blocks by selecting the output port (with symbol >) and moving it to the
input port (also with symbol >). Once the connection is made, a connecting line shows up,
and the input and output ports disappear.
3. Double-click a block to assign numerical values or specifications if needed.
4. Add the Scope block from the Sinks library for outputting the simulation results
on a selected variable.

These actions don’t have to follow the order listed above, and they can be taken alternatively
at the user’s convenience.
There are two points regarding this process of model building. First, because Simulink is
MATLAB-based, some blocks, such as Gain and Integrator, have default numerical
values to begin with. These values often need to be changed according to the problem in
consideration. To do this, double-click the block and assign the desired value(s). Second,
when constructing a model, the user is recommended to follow the five-step procedure
presented in Section 6.5. This is simply because the formation of a time-domain block
diagram is quite similar to that for a Simulink model.

Simulation with a Model

A simulation with a Simulink model can be run in one of the following two ways:

• Click on the Run icon on the toolbar of the model window

• In the MATLAB Command Window, call the model at the prompt by typing sim
(‘model_name’), where model_name is the file name of the model.

On the toolbar of the model window, the user can specify a desired stop time for simulation.
The default stop time is 10.0. After the simulation is executed, double-clicking the Scope
blocks will show the simulation results.
As a demonstrative example, consider a spring–mass–damper system given by

x þ cx_ þ kx ¼ q
m€

with m = 2, c = 3, and k = 32. Assume a constant external force (input), qðtÞ ¼ 16 uðtÞ, where
uðtÞ is a unit step function. The initial conditions of the system are given as follows
_
xð0Þ ¼ 0:2; xð0Þ ¼ 1. Let the system outputs be the displacement x and the damping
_ A Simulink model for the system is built to compute the system outputs for
force fd ¼ cx.
0 ≤ t ≤ 8:0.
According to the time-domain block diagram of the system shown in Figure 6.5.2, we
obtain the corresponding block components by the Library Browser and place them in a
model window (saved with name SMD_Model); see Figure B2.3, where the far-left block is
the Step block from the Sources library.
 Appendix B 767

Figure B2.3 Blocks picked up for block diagram construction

Connecting all the blocks in the model window, naming the blocks, double-clicking each
block for updates, and assigning the parameter values of the Gain blocks gives an incom-
plete block diagram shown in Figure B2.4.
Finally, adding a Gain block and two Scope blocks for the system outputs completes the
construction of the block diagram for the spring–mass–damper system; see Figure B2.5.
Before running a simulation, three things need to be done. First, set up the stop time as 8.0
on the toolbar. Second, assign the initial conditions of the system, which is done by double-
clicking each of the integrators and assigning an initial value (Figure B2.6). Third, for a small
enough step size in the simulation, click on the Configuration Parameters icon on the
toolbar of the MODELING tab and, in the open window, choose the Fixed-step for Solver
selection and set the Fixed-step size to a small number, say 0.005. Afterwards, save all
changes made on the model.
Now, run the Simulink model. Once the job is done, double-click the scopes to see the
system outputs in Figure B2.7. Here, in Figure B2.7(a), the x plot is obtained by the
command Copy to Clipboard from the File menu of the Scope, and the curve in Figure
B2.7(a) is an original plot of fd from the Scope block.
768 Appendix B

Figure B2.4 Connected blocks of an incomplete block diagram

Figure B2.5 Completed block diagram of the spring-mass-damper system

 Appendix B 769

Figure B2.6 Completed block diagram for the spring–mass–damper system

770 Appendix B

(a)

0.8

0.6

0.4
x

0.2

−0.2

0 1 2 3 4 5 6 7 8

(b)

Figure B2.7 The outputs of the spring–mass–damper system: (a) displacement x; and (b) damping force fd
 APPENDIX C
A Brief Introduction to Wolfram Mathematica

This appendix gives a brief introduction to Wolfram Mathematica. A detailed tutorial about
these software tools is available on the publisher’s website and through Wolfram Research
Institute documentation.
Wolfram Mathematica is a state-of-the art software tool for the widest variety of symbolic
and numerical computations, as well as data visualization and analysis. Among scientists,
Mathematica has been long renowned as the ultimate computation platform due to its
capacity for superior symbolic manipulations, numeric computations, data analysis, graph-
ics, and visualization.
Wolfram Language – the programming language of Mathematica – is a functional lan-
guage, which allows for massive parallelization of computations. It is a knowledge-based
interpreted language, with the scripts (notebooks) evaluating from top to bottom. Well-
designed intuitive interfaces provide for a high degree of interactivity and programming ease.
A vast library of built-in functions (version 12 of Mathematica has 6981 functions),
a boundless capacity for extending the built-in capabilities with custom functions, and the
high level of abstraction of the code allow for the fast development of relatively large projects
by small teams.

C.1 Mathematica Operators and Functions

Mathematica uses a “notebook” interface to receive input and produce symbolic,
numerical, graphical, or interactive output. When the software is launched, an empty
notebook with a blinking cursor is displayed. Useful data-entering and notebook-
formatting assistants – palettes – allow for typing equations just like they are written
on paper. These palettes are available from the drop-down menu as shown in Figure
C1 below (one of the most popular palettes – Basic Math Input – is shown at the left of
the notebook).
Listed in Tables C1 to C9 are some commonly use Mathematica functions. Typing ?
FunctionName in the notebook will display what the function named FunctionName
is, and how to use it. If the name of the sought function is not known, Mathematica
help allows to type a partial name, in response to which it will yield a list of possible
matches.

771
772 Appendix C

Figure C1 Mathematica palette: drop-down menu and palette

Table C1 Operators and special characters

+ Addition = Assignment operator

‒ Subtraction > Less than
* Multiplication ≤ Less than or equal to
/ Division > Greater than
^ Power ≥ Greater than or equal to
⋅ Matrix multiplication operator == Equal to
(* *) Code comment != Not equal to
{} indicates a list (non-scalar variable) && Logical operator: AND
[] argument of a function || Logical operator: OR
[[ ]] accessing list element by address ! Logical operator: NOT
 Appendix C 773

Table C2 Commonly used mathematical functions of Mathematica (x indicates scalar

argument, xList indicates list argument)

Function Description

Sqrt [x] Square root

Surd [x, n] Real-valued n-th root of x; if x is symbolic, it is assumed real-valued
Exp [x] Exponential
Log [x] Natural logarithm (base-e)
Log10 [x] Base-10 logarithm
Sign [x] Signum function
Abs [x] Absolute value
Round [x] Yields the integer closest to the value of x
Floor [x] Rounds x to the closest integer n ≤ x
Ceiling [x] Rounds x to the closest integer n ≥ x
Mod [x, y] Yields the remainder of division of x by y
Min [xList] Minimum value of xList
Max [xList] Maximum value of xList
Mean [xList] Statistical mean of xList
Sum [xList] Sum of elements of xList
Re [x] Complex real part
Im [x] Complex imaginary part
Arg [x] Phase angle in radians
Conjugate [x] Complex conjugate
ComplexExpand [x] Expands symbolic expression x into real and imaginary parts
ExpToTrig [x] Converts complex exponential x into trigonometric functions using
Euler formula
TrigToExp [x] Converts a trigonometric expression x into a sum of complex
exponentials using Euler formula
Apart [x] Decomposes a rational expression into partial fractions
Together [x] Takes a sum of fractions to a common denominator, canceling factors in
the result
Collect [expr, x] Collects together terms in expression expr involving the same powers of
objects matching x
Sin [x] Sine
Cos [x] Cosine
Tan [x] Tangent
ArcSin [x] Arcsine
ArcCos [x] Arccosine
ArcTan [x] Arctangent
Sinh [x] Hyperbolic sine
Cosh [x] Hyperbolic cosine
Tanh [x] Hyperbolic tangent
774 Appendix C

Table C3 Matrix functions (xMat argument is a two-dimensional list that is a full array)

Function Utility

Det [xMat] Determinant

DiagonalMatrix [xList] Creates a diagonal matrix with the elements of xList
on the main diagonal
DiagonalMatrix [xList,k] Creates a diagonal matrix with the elements of xList
on the k-th diagonal
IdentityMatrix [n] Creates an identity matrix of the size n  n
Dot Scalar product of matrices, automatically handling
row and column vectors
Cross [u, v] Vector (cross) product of two vectors u and v
VectorAngle [u, v] Angle between vectors u and v
Transpose [xMat] Matrix transpose
Inverse [xMat] Matrix inverse
MatrixPower [xMat] Power of numeric or symbolic matrix
MatrixExp [xMat] Matrix exponential
MatrixLog [xMat] Matrix natural logarithm
Eigensystem [xMat] Generates a list of eigenvalues and eigenvectors
Eigenvalues [xMat] Generates a list of eigenvalues
Eigenvectors [xMat] Generates a list of eigenvectors
Norm [xMat] Matrix and vector norm
Normalize [xMat] Normalize vector xMat to unit length

Table C4 Graphics generation and styling functions

Function Utility

Plot [f, {x,xmin,xmax}] Generates a plot of a single function f ð xÞ on the interval

x 2 ½xmin ; xmax 
Plot [{f1,f2,. . .,fn}, Generates a plot of n functions fi of an independent
{x,xmin,xmax}] variable x on the interval x 2 ½xmin ; xmax 
ParametricPlot [{fx,fy}, Generates a plot of a single function f defined paramet-
{t,tmin,tmax}] rically, where its x- and y-components fx ðtÞ and fy ðtÞ
are functions of an independent variable t, on the
interval t 2 ½tmin ; tmax 
ParametricPlot [{{f1x,f1y}, Generates a plot of n parametric functions fi , where the
{f2x,f2y},. . ., {fnx,fny}}, interval for the independent variable is t 2 ½tmin ; tmax 
{t,tmin,tmax}]
PolarPlot [r,{θ,θmin,θmax}] Generates a polar plot of a single function rðθÞ, where r is
the polar radius and θ is the polar angle, on the inter-
val θ 2 ½θmin ; θmax 
LogPlot [f,{x,xmin,xmax}] Generates a linear-logarithmic plot of a function f ð xÞ,
where x-axis is linearly scaled and y-axis is logarith-
mically scaled, on the interval x 2 ½xmin ; xmax 
 Appendix C 775

Table C4 (cont.)

Function Utility
LogLinearPlot [f,{x,xmin,xmax}] Generates a log-linear plot of a function f ðxÞ, where x-
axis is logarithmically scaled and y-axis is linearly
scaled, on the interval x 2 ½xmin ; xmax 
LogLogPlot [f,{x,xmin,xmax}] Generates a log-log plot of a function f ð xÞ, where both
axes are logarithmically scaled, on the interval
x 2 ½xmin ; xmax 
Graphics [g_primitives, Generates styled graphic primitives such as circles,
styling_opts] polygons, lines, etc.
Plot3D [f,{x,xmin,xmax}, Generates a plot of a function z ¼ f ð x; yÞ on the intervals
{y,ymin,ymax}] for its independent variables x 2 ½xmin ; xmax  and
y 2 ½ymin ; ymax 
Plot3D [{f1,f2,. . .,fn}, Generates a plot of n functions fi of two independent
{x,xmin,xmax}, {y,ymin,ymax}] variables x and y on the interval x 2 ½xmin ; xmax  and
y 2 ½ymin ; ymax 
ParametricPlot3D [{fx,fy,fz}, Generates a plot of function f defined parametrically,
{t,tmin,tmax}] where its x-, y-, and z- components fx ðtÞ, fy ðtÞ, and fz ðtÞ
are functions of an independent variable t, on the
interval t 2 ½tmin ; tmax 
Graphics3D [g_primitives, Generates styled three-dimensional graphic primitives
styling_opts] such as spheres, cylinders, polyhedrons, etc.
Plot styling options
PlotRange Specifies the range of coordinates to be included in the
generated plot
PlotStyle Specifies styles in which plot elements need to be
generated; common directives apply to line color, type
(continuous, dashed, etc.), thickness, surface color
and opacity, and point size
PlotLegends For multiple graphs on the same plot specifies the labels
for each graph and legend location on the plot figure
PlotLabel Specifies plot title and its appearance
Axes Specifies whether axes need to be drawn
AxesOrigin Specifies where the drawn axes need to cross
AxesLabel Specifies labels for the drawn axes
AxesEdge For three-dimensional graphics specifies on which edges
of the bounding box to draw the axes
Frame Specifies whether to draw a frame around the two-
dimensional plot object
FrameLabel Specifies whether labels need to be placed on the edges of
a frame, and on which edges they need to be
GridLines For two-dimensional graphics specifies whether to draw
the grid lines, and grid line spacing in x- and y- directions
FaceGrids For three-dimensional graphics specifies on which faces
of the bounding box to draw the grid lines, and grid
line spacing in x-, y-, and z- directions
776 Appendix C

Table C5 Equation solvers

Function Description

Solve [eqs,vars] Solves a system eqs of algebraic equations or inequalities for the
variables vars; yields a closed-form solution
NSolve [eqs,vars] Finds numerical approximations to the solutions of the system eqs
of algebraic equations or inequalities for the variables vars
FindRoot [eqs,x0] Finds numerically local solutions of the system eqs of algebraic
equations, starting from the point x ¼ x0
FindInstance [eqs, vars] Finds numerically an instance of the variable vars values that satisfy
the system eqs of algebraic equations
LinearSolve Solves a system of linear algebraic equations in matrix form
Roots Finds roots of a single polynomial
DSolve Finds exact solutions to differential, delay, and hybrid equations
DSolveValue
NDSolve Finds numerical approximations to the solutions to linear and
NDSolveValue nonlinear ordinary differential equations (ODEs), partial
differential equations (PDEs), differential algebraic equations
(DAEs), delay differential equations (DDEs), integral equations,
integro-differential equations, and hybrid differential equations
ParametricNDSolve Finds numerical approximation to the solution to ordinary and
partial differential equations with parameters

Table C6 Transfer function formulation

Function Description

TransferFunctionModel Creates a transfer function model;

can take a linear state-space model as an argument to
generate the transfer function representation
TransferFunctionPoles Gives a matrix of poles of the transfer function
TransferFunctionZeros Gives a matrix of zeros of the transfer function
SystemsModelSeriesConnect Series interconnection of the two blocks (transfer
functions)
SystemModelParallelConnect Parallel interconnection of the two blocks (transfer
functions)
SystemModelFeedbackConnect Obtains the transfer function of a feedback control
system
 Appendix C 777

Table C7 State-space formulation

Function Description

StateSpaceModel Creates a linear state-space model;

can take a transfer function model as an argument
to generate the state-space representation
NonlinearStateSpaceModel Creates a nonlinear state-space model

Table C8 Response

Function Description

StateResponse Gives the numeric or symbolic state response of the state-space

model to a given input
OutputResponse Gives the numeric or symbolic output response of the state-space
model to a given input
Note: to view a symbolic response in its simplest form, applying functions Simplify and
FullSimplify to the generated response expression is recommended.

Table C9 Classical analysis and design

Function Description

RootLocusPlot Plot root locations as a parameter varies

BodePlot Bode plot (magnitude and phase) of the frequency response
NyquistPlot Nyquist plot (polar plot of the frequency response)
GainPhaseMargins Gain and phase margins
PIDTune Automatic PID tuning for any linear systems
 Index

accelerometers 730 back emf, DC motor 439–40, 444 Bode diagram/plot 562–3
capacitive 734–5 balanced force micro-accelerometer first-order system 564–6
electromechanical 730, 732–3 734–5 second-order system 568–71
piezoelectric 735–42 band-pass passive filter 274–7 bounded input–bounded output
vibrometer and 733–4 bandwidth-limited differentiator (BIBO) 545–6
active-circuit analysis 277 circuit 289 branches 623–4
isolation amplifiers 277–81 bandwidth-limited integrator circuit branch point 414–15
operational amplifiers 281–97 287, 288 breakaway point 625, 627
active elements 223, 230 Bernoulli equation 370–4 break frequency 565, 569
active microphones 743 bimetallic strip thermometer 697 break-in point 625, 627
actuator 353, 590, 592 copper–steel and aluminum–
adder circuit 291, 292 tungsten 701–3 cantilever-beam accelerometer 737–8
adjoint matrix 30–1 dynamic modeling 698–701 capacitive accelerometers 734–5
algebraic manipulations 18–19, 36 schematic and lumped-parameter capacitors 227–8
alternating current (AC) 433, 558 model 699 Cauchy form 401
aluminum–tungsten thermometer 701–2 spiral 697, 698 cause-and-effect approach 414
analog devices 433 Biot number 345–6 closed-loop control system 590–2
analogous relations block diagrams 396, 398–9, 401 advantages 591
element laws 12–13 armature-controlled DC-motor 441–2 block diagrams 592
flow and effort 11–12 branch/takeoff point 414 configuration 591
variables and equivalent coefficients closed-loop 592 coupled engine–propeller system
13, 14 DC motor with load 455 695–7
analysis, physical systems 4 dynamic system with 417 disturbance rejection 598–9
anti-commutative property, cross electrical circuit 258–65 output of 600
product 22–3 equivalent transformation 415–16, parameter variations 596–8
armature 431–4 418–19, 423–4 PID gains effect 606
armature-controlled DC motors 436, 438 expanded 441–2, 445–6 root locus of 621–2
angular velocity 440 field-controlled DC-motor 446 speed control, DC motor 595
block diagram 441 linear time-invariant dynamic system step response 609
electrical subsystem 439 414–15 system response improvement 599–601
expanded block diagram 441–2 mechanical systems 181–5 commutator 435–6
Kirchhoff’s voltage law 439 open-loop 592, 688 complementary homogeneous
with load 448, 452–3 parts of 462–3 equation 80–2
mechanical subsystem 439 PID controller 602 complete solution, linear differential
state-space representation 442–3 in s-domain 9–10, 182, 457, 461 equation 91–2
transfer function 441 second-order system 420–1 complex numbers
asymptotes 623 signal distribution 417–18 algebraic manipulations 36
angles vs. n − m 626 summation point 414 complex conjugates 35–6
root locus 624–5 system transfer function 418–19 definition 34
asymptotically stable system 545–6 in time domain 10–11, 457–69 magnitude and phase angle 34–5
automatic control system 589 time-invariant systems 97–8 phase angles, quadrants 35

778
 Index 779

trigonometric manipulations 36–7 damping elements, rotational systems dry friction 125–7
computation software 13, 15 149–51 dynamic modeling, half-car model
condenser microphone 753–6 damping force 123, 125, 770 669–70
conduction 336, 337 damping ratio 518, 521, 532, 567–8, 571 Lagrangian approach 672–3
conservative force 119 damping states, second-order system Newtonian approach 670–2
constant-pressure process 355 517–18 dynamic responses
constant-temperature process 356 D’Arsonval meter/movement 430–3 Mathematica 198–203, 297–306
constant-volume process 356 DC electrical circuit 225 MATLAB 191–8, 378–83
continuous-time models vs. discrete- DC motors. See direct current (DC) dynamic systems
time models 5 motors analogous relations 11–13
controlled mechanical system 399–401 degrees of freedom (DOFs) 160, 179, application problems 666–756
control logic 590 667, 669 classification 4–5
control systems 589 diagonal matrix 24 combined systems and system
actuator 590 differential equations modeling techniques 396–485
closed-loop 590–2 applications 63 components 1–2
controller 590 linear and nonlinear 63–4 computation software 13, 15
liquid-level 652–3 linear ordinary 70–95 electrical systems 222–306
open-loop 590 partial 66–7 feedback control systems 589–653
PID gain tuning 638–9 spring–mass–damper dynamic input and output 2–3
plant 589–90 system 64–5 mathematics fundamentals 16–102
robotic system 648–50 state equations 67–9 mechanical systems 114–203
robustness of 596 two-mass dynamic system 65–6 system model representations 6–11
room-temperature 591–2 See also specific differential system response analysis 502–77
sensor 590 equations thermal and fluid systems 334–83
stability analysis 613–15 differentiator circuits 288–90 units 15
steady-state error 615–19 dipolar-charge electrets 735
steady-state response 611–13 direct current (DC) motors 396 electret(s)
thermal 650–2 armature-controlled 438–43 -based microphones 755–6
transfer function formulation 592–5 bearings 434 dipolar-charge 735
transient response 611–13 electrical subsystems 437–9 piezoelectric crystal 736
by zero placement 639–48 elements 434 real-charge 735
convection 337, 341 field-controlled 443–7 electrical circuit 223. See also electrical
convolution theorem 52–3 with load 447–56 systems
copper–steel thermometer 701–2 mechanical subsystems 437–9 electrical subsystem 397–8, 425,
corner frequency 565, 569 modeling 436–8 442
coupled engine–propeller system 686–7 operations 433–6 condenser microphone 754
closed-loop system 695–7 rotor 434 DC motor 437–9
open-loop system 687–95 shaft 434 dynamic microphone 744
coupling relations 397 simplification 456–7 field-controlled DC motor 443–5
cover-up method 54 stator 434 galvanometer 432
Cramer’s rule 28–30, 144 torque–current relations 437–8 loudspeaker 746
critically damped system 517, 520, 523, direct integration 71–2 open-loop system 687, 688
525 first-order differential equation 72 electrical systems 222–3
crossover points 627–8 second-order differential equation 73 active-circuit analysis 277–97
cross product 21–3 discrete-time models 5 block diagrams 258–65
current divider rule 237–9 displacement transmissibility 570, 571 current and voltage 224–5
current element 426–7 distributed models vs. lumped models 4 definition 223
current-isolation amplifiers 279 DOFs. See degrees of freedom (DOFs) fundamentals 223–30
current-to-voltage converter 296 dominant poles 543–4, 635 impedance 230–41
cutoff frequency 271, 272 dot product 21–3 Kirchhoff’s laws 242–4
780 Index

electrical systems (cont.) separation of variables 73–4 Bode diagram/plot 562–3

Mathematica, dynamic responses first-order high-pass filter 289 first-order systems 564–6
297–306 first-order systems 457 general concepts 559–61
passive-circuit analysis 244–57 applications 508 Nyquist plot 563–4
passive filters 271–7 Bode diagram 564–6 plots 561–4
state-space representation 265–70 forced response 509–10 second-order systems 567–71
electroacoustic devices 742–56 free response 509 utilities 560
electromagnetism 425–30 frequency response 564–6 friction forces 125–7, 150–1
electromechanical systems 396, 425 general inputs, response to 512–13
D’Arsonval meter 430–3 impulse response 510–11 galvanometer 430–3
electromagnetism 425–30 PID controller 603 geared systems 162, 164–8
electromotive force (emf) 429 ramp response 511–12 kinematic relation 163
electronic cigarettes 702 step response 511 spur gears 162–3
energy conservation 338–9 time constant estimation 513–14 torque ratio 163–4
energy functions 179 fluid capacitance 356, 358–9 work done by torques 164
kinetic energy 177–8 fluid inertance 353, 356–7, 368–70 general differential equations 90
potential energy 119, 178 fluid resistance 356, 360–2
Rayleigh dissipation function 178 fluid systems half-car model 667, 669–71
equations of motion derivation duct/pipe 354–5 heat energy 336, 338
rotational systems 145–53 gases, equation of state 355–6 heating elements 704
translational systems 130–8 hydraulic systems 353 governing equations 704–6
error signal 416 mass conservation 354 properties of 704
Euler’s formula 34, 37, 44, 520 pneumatic systems 353 stainless-steel 704
Evans, W. R. 619 properties 334 temperature change and elongation
exponential function 43–4, 47 SI units 759 706–12
See also liquid-level systems heat transfer
Faraday’s law of induction 429, 744 forced response 70–1 conduction 336, 337
feedback control systems 590 boundedness 550 convection 337
root locus 619–48 first-order systems 509–10 radiation 337–8
rotor system 686–7 second-order system 522 thermal resistance 340–1
Simulink model 611 stability by 545 higher-order systems, time responses
step response 637, 642, 646 system response analysis 503–5 539–44
time response 613, 615 forces and moments 116–17 dominant poles 543–4
by zero placement 640–1 force transmissibility 570–1 general 541–2
See also closed-loop control system; free-body diagrams and auxiliary plots model reduction 543–4
control systems equations of motion derivation 130– third-order systems 540–1
ferroelectrics 736 6 high-fidelity audio systems 743
field-controlled DC motor 436 motor–propeller assembly 151–3 high-pass filters 272–4, 275
block diagram 446 rigid-body system 172 Hooke’s law 121, 728
with load 450 rotational systems 151–3 hydraulic systems 353, 370
mechanical and electrical subsystems translational systems 127–9
443–5 free response 70–1 ideal current source 223
state-space representation 446–7 damping cases 520–2 ideal gas law 355
final-value theorem 52, 507–8 first-order systems 509 ideal voltage source 223
finite-dimensional system 4 stability by 545–6 identity matrix 24
first-order differential equations system response analysis 503–5 imaginary axis crossings 627
direct integration 72 frequency domain, system response impedance
particular solutions 82–4 analysis 558–71 circuit analysis 231
piecewise-continuous forcing frequency response 558 current divider rule 237–9
function 78–9 advantages 558 definition 230–1
 Index 781

differentiator op-amp circuit 288, 289 lag circuit 287 linear time-invariant systems 398
equivalent form 231, 233–5 lag/firstorder low-pass filter 287 block diagram 97, 414–15
integrator op-amp circuit 285, 287 lag-lead op-amp circuit 290 Laplace transform for 398
inverter op-amp circuit 285 Lagrange’s equations 127, 179–81 spring–mass–damper system in 460–
lead-lag circuit 290–1 Lagrangian approach 177, 670, 672–3 1
low-pass filter 271 laminar flow 361 stability states of 545–6
op-amp 281–3 Laplace transform 37, 398, 440 steady-state response 414, 558
parallel connection 232–3 convolution theorem 52–3 time response 571–7
series connection 231–2 definition 38–9 liquid-level systems
voltage divider rule 235–7, 239–41 differentiation (derivative property) dynamic modeling 363–8
voltage-isolation amplifier 278–9 48–9 energy sources 363
impulse response evaluation 42 engineering applications 334
damping cases 522–4 exponential function 43–4 flow source 363
first-order systems 510–11 factoring 42 fluid capacitance 358–9
mechanical systems 549 final-value theorem 52 fluid resistance 360–2
inductors 228–9 functions, derivation 46–8 hydraulic systems 356
inertia effect 368 governing equations 409, 412, 421 mass conservation 359–60
infinite-dimensional system 4 initial-value theorem 51–2 orifice resistance formulas 362–3
initial-value theorem 51–2 integration (integral property) 49–50 pressure source 363
input–output differential-equations 399 linearity property 42 storage tanks 357
input–output relations 2–3, 258 multiplication by exponential and t water purification 711–23
integrator circuits 285–8 46 loading effect 281
International System of Units (SI) 15 pairs 39, 40 loop currents 245
base units 758 piecewise-continuous functions 60–2 loop method 245–7, 253
conversion factors 760 poles and zeros, function F (s) 39, 41 Lorentz’s force law 426, 429
electrical and magnetic systems 759 ramp function 43 loudspeaker 743
fluid capacitance 358 sinusoidal function 44 actual system vs. second-order
mechanical systems 758–9 step function 43 approximation 750–1
thermal and fluid systems 759–60 time-shifted function 44–5 block diagram 747
thermal capacitance 339 lead circuit 289, 290 dynamic response 748–9
inverse Laplace transform 38 lead-lag op-amp circuit 290 input voltage 749
imaginary and complex poles 58–9 Lenz’s law 429 mechanical and electrical subsystems
partial fraction expansion 53–8 lever–spring system 155–6 746
time-domain response 76 linear differential equation 63–4 original system vs. second-order
inverse piezoelectric effect 736 linear integro-differential expression system 751, 752
inverter (sign-changer) circuits 285 50–1 piecewise-continuous input 749, 750
invertible matrix 30 Linear models vs. nonlinear models 4–5 simple intercom assembly 753
isentropic process 356 linear ordinary differential equations transfer function 749–50
isobaric process 355 direct integration 71–3 low-pass filters 271–3
isochoric process 356 free response and forced response lumped masses 120–1, 175
isolation amplifiers 277–81 70–1
non-inverting op-amps 294–5 Laplace transform 74–9 magnetic field strength 425
isothermal process 356 method of undetermined coefficients magnetic force 426–8
80–92 manual PID tuning 606
kinetic energy 177–8 piecewise-continuous forcing marginally stable system 546, 548
Kirchhoff’s current law (KCL) 242, functions 93–5 mass elements 147, 172
247, 283, 284, 291, 293 separation of variables 73–4 MATLAB 3, 99–100, 397, 573–6, 761
Kirchhoff’s voltage law (KVL) 242, linear resistor 227 bode diagrams 562
245, 258, 271, 276, 278, 279, 432, linear state-space representations 402– dynamic responses 191–8, 378–83
439, 444 3, 408 equation solvers 763
782 Index

MATLAB (cont.) loudspeaker 746–53 output of 600

frequency response 764 mechanical and electrical subsystems parameter variations 690–2
function blocks, nonlinear state 744–6 parts of 687–8
equations 576–7 parts 743 speed control, DC motor 595
functions for graphics 763 passive and active 743 open-loop poles/zeros 621, 623
linear/nonlinear systems, time sound waves 743–4 operational amplifiers (op-amps)
response 573–7 velocity-sensitive 746 circuits
mathematical functions 762 voltage output 745 adder, subtractor and comparator
matrix functions 762 Modelica System Model 481–2 291–2
Nyquist plots 563 modeling and simulation 3 amplification factor 282
operators and special characters 762 by MATLAB/Simulink 464–75 converters 295–6
root locus 628, 631–2 by Wolfram Mathematica 475–85 differentiator 288–90
state-space formulation 763 model parameters 515, 535 ideal inverting 281
time response of car 683–6 graphic interpretation 529 integrator 285–8
transfer function formulation 763 moment of force 116–17 integrator–differentiator
See also Simulink monocrystalline 736 combination 290–1
matrix algebra motor–propeller assembly 152–3, 686–7 inverter (sign-changer) 285
definition 24 multiple inputs and multiple outputs with load 296–7
determinants 25–30 (MIMO systems) 2, 411, 414, 421 multiplier (amplifier) 283–5
inverse of 30–1 multiplier (amplifier) circuits 283–5 non-inverting amplifiers 293–5
operations 32–3 non-inverting input 281, 283
single-matrix operations 24–5 NDSolve functionality 99, 100 typical inverting 282
matrix determinants 25–30 net currents 245 ordinary differential equations (ODEs)
matrix inverse 30–1 Newtonian approach 177, 670–2 4
maximum overshoot 530–2 Newton’s laws of motion 115–16 orifice resistance 362–3
mechanical subsystem 397–401, 425, Nichols, N. B. 606 overdamped system 517, 520–1, 523,
442 node method 247–9 525
DC motor 437–9 node voltages 247
DC motor with load 447 non-conservative force 119 parallel-axis theorem 146, 147
field-controlled DC motor 443–5 non-inverting op-amp circuits 293–5 parallel connection, impedance 232–3
open-loop system 687, 688 nonlinear air damping effect 692–5 partial differential equations (PDEs) 4,
mechanical systems 114 nonlinear differential equation 63 66–7
block diagrams 181–5 nonlinear state equations, MATLAB partial fraction expansion, inverse
energy approach 177–81 576–7 Laplace transform 53–8
fundamental principles 115–20 nonlinear systems, time response 571–7 particle kinematics 117–18
Mathematica 198–203 nonsingular matrix 30 particular solution 80, 82
MATLAB 191–8 Norton’s theorem 251, 255–6 first-order differential equations 82–4
rigid-body systems, plane motion null matrix 24 second-order differential equations
169–77 numerical integration method 572 85–91
rotational systems 144–68 Nyquist plot 563–4 passive-circuit analysis 244
state-space representations 185–91 loop method 245–7, 253
translational systems 120–44 Ohm’s law 226, 227, 245, 247, 251, 278, node method 247–9
mechanical workhorse 434 291, 293 Norton’s theorem 251, 255–6
metric system 15 open-loop control system 590 superposition theorem 249–50, 256–7
micro-electromechanical systems block diagrams 592, 688 Thevenin’s theorem 250–1, 253–5
(MEMS) accelerometers 734, 737 configuration 591 passive elements 223
microphones, dynamic 743 coupled engine–propeller system passive filters 271
condenser 753–6 687–95 band-pass filter 274–7
current output 746 ideal case 688–90 high-pass filter 272–4, 275
fidelity 743 nonlinear air damping effect 692–5 low-pass filter 271–2, 273
 Index 783

passive microphones 743 Ziegler–Nichols gain tuning 606–10 magnitude criteria 623
peak time 530, 531 pulleys 153–4 vs. parameter k 620–1
performance specification parameters plot 620–1
529–36 quarter-car suspension model 133–5, properties 622–8
permanent-magnet moving-coil 142, 179, 667 real-axis segments 623–4
movement 431 quasi-static deflections of car 675–6 sketching 628–32
phase angles 34, 35, 529, 531, 569, 677 stability analysis 632–5
PID controller. See proportional– radiation 337–8, 341 system response analysis 635–8
integral–derivative (PID) controller ramp function 43 rotating shafts 150, 151, 157
PID feedback control law 50 ramp response, first-order systems rotational systems 144
piecewise-continuous forcing functions 511–12 damping elements 149–51
93–5 Rayleigh dissipation function 178 double-headed arrows 145
piecewise-continuous functions 60–2 real-charge electrets 735 free-body diagrams and auxiliary
piezoelectric accelerometers 738–42 rectangular pulse, piecewise- plots 151–3
construction of 737 continuous function 61–2 geared systems 162–8
crystals 735–7 resistive-heating elements 702. See also lever 155–6
electrets 735–6 heating elements mass elements 147
ferroelectrics 736 resistors 226–7 parallel-axis theorem 146, 147
inverse piezoelectric effect 736 resonant frequency 567–8 principle of impulse and momentum
moderate-to-low cost 737 resonant response 538 147
monocrystalline materials 736 resonant vibration 85–8 principle of work and energy 146–7
Young’s modulus 736–7 revisable adiabatic process 356 pulleys 153–4
piezoelectric effect 736, 737 right-hand rule 21–3, 145, 426, 427, 430 rigid bodies 145–6
piezoresistive accelerometers 737–8 rigid-body systems, plane motion rotating shafts 157
plant 397, 589–90 free-body diagrams and auxiliary simple pendulum 154–5
pneumatic systems 353 plots 172 slewing rigid bar 156–7
basic elements 374–6 fundamental principle 171–2 spinning rigid disk 157
fundamental principle 374 lumped masses and 175–7 spring elements 147–9
modeling 376–8 mass center and relative motion transfer function formulation 168
polytropic process 356 169–71 translational elements and 158–62
potential energy 119, 178, 357 mass elements 172 rotor system 686–90, 695–6
potentiometers 239–40 pulleys/disks with movable pins nonlinear air damping 692–3
principle of impulse and momentum 172–3 parameters of 689
120, 147 spring and damping elements 172 simulink model 694
principle of work and energy 118–19, wheel in rolling and sliding motion Routh array 551–4, 556
146–7 173–5 Routh–Hurwitz stability criterion
proportional–integral–derivative (PID) rise time 530, 531 550–1, 614
controller 601–6 robotic system 648–50 application 552, 554–5
algorithm 601–11 root locus method 619 features 551
block diagram 602 analysis and design 632–48 necessary condition 551–2
first-order system 603 angle criteria 623 Routh array 552–4
gain tuning 606–10, 638–9 asymptotes 623–6 rules 553
manual tuning 606 branches 623–4 system orders two to four 555–8
proportional (P) 602 breakaway point 625, 627 Runge–Kutta method 99, 101, 573,
proportional plus derivative (PD) break-in point 625, 627 682–3
593–4, 603 closed-loop system 621–2
proportional plus integral (PI) 603 concepts 620–2 scalar product 21
Simulink, system responses 611 control system design 638–48 s-domain block diagrams 9–10, 182,
step-response curves 605 crossover points 627–8 457, 461
tuning 606 feedback system 634–5 second-order band-pass filter 290–1
784 Index

second-order differential equations 49, undamped systems 537–8 steady-state vibration of car 676–81
51, 291, 445 SI units. See International System of step function 43
direct integration 73 Units (SI) step response
Laplace transform 76–7 slewing rigid bar 156–7 closed-loop system 609
particular solutions 85–91 specific heat capacity 338 curves PID controller 605
second-order systems spinning rigid disk 157 damping cases 525–6
damping cases 517–19 spring elements 147–9, 172 feedback system 637
free response 520–2 spring–mass–damper systems 4, 6, 95, first-order systems 511
frequency response 567–71 571 third-order systems 544
general form 526–8 block diagram of 9, 97, 767–9 underdamped second-order systems
impulse response 522–4 capacitance, resistance, and 528–37
sinusoidal input to 538–9 inertance 12–13 storage tanks 357
step response 525–6 free-body diagrams 128, 129 subtractor circuit 292
time responses 515–28, 537–9 free response and forced response 71 summation point 414
underdamped 528–37 mechanical systems and fluid systems superposition theorem 249–50, 256–7
seismograph 733 13, 14 system analysis 398
sensors 590 nonlinear differential equation 11 system-level modeling 396–8
accelerometers 730, 732–5 second-order linear differential system modeling techniques 398–401
definition 723 equation 64–5 block diagrams 414–24
displacement, sinusoidal motion second-order systems 515 governing equations 399
726–31 state variables 8 input–output differential-equations
strain gauge 724, 726 transfer function 7, 95 399
Wheatstone bridge 724 spur gears 162–3 state-space representations 401–8
series connection, impedance 231–2 stability analysis 544–5 transfer function formulations 408–13
settling time 530–4 conditions in pole locations 546–50 system model representations
simple pendulum 154–5 definitions 545–6 s-domain block diagram 9–10
Simulink 3, 378–83, 464 feedback control system 613–15 state-space representation 7–9
closed-loop system 695 root locus method 632–5 time domain, block diagram in 10–11
DC motor 471–3 Routh–Hurwitz stability criterion transfer function formulation 6–7
feedback control system 611 550–8 system order and input type 508
MATLAB function blocks, states of 545–6
nonlinear state equations 576–7 systems of order 555–8 takeoff point 414
model building 765–6 stable system 546–8 thermal control system 650–2
model window and Library Browser state equations 67–9, 186 thermal systems
761, 764–5 linear and nonlinear 401 Biot number 345–6
nonlinear system 575–6 numerical integration 98–102, 572 energy conservation 338–9
open-loop system 693–4 state-space representations 7–9, 399, 414 engineering applications 334
simulation 766–70 differential equation 405–7 free-body diagrams and auxiliary
solution by 573–6 dynamic system 402 plots 346–53
system responses by 611 electrical systems 265–70 heat energy 338
on time-domain block diagrams 465–9 mechanical systems 185–91 heat transfer 336–8
transfer functions and state-space state/output equations 401–2 linear lumped-parameter models 335
blocks 469–75 vibration analysis of car 673–5 MATLAB and Simulink, dynamic
single-input single-output (SISO) stator 434 responses 378–83
system 2 steady-state error 615–19 properties 334
single-matrix operations 24–5 steady-state response SI units 759
sinusoidal function 44, 668 final-value theorem 507–8 thermal capacitance 339–40
sinusoidal input, second-order systems frequency response 560 thermal energy 335–6
537 linear time-invariant system 558 thermal resistance 340–5
damped system 538–9 transient response and 505–7 Thevenin’s theorem 250–1, 253–5
 Index 785

third-order differential equation 77–8 time-invariant systems 95–7, 139 unity-gain buffer 294
third-order systems 540–1 for two-input–two-output system unstable system 546–8
three-degrees-of-freedom (3-DOF) 411–13
model 667, 669 transfer matrix 411 vector algebra 16–17
three-term controller 601 transient response 505–7 addition and subtraction 18
time constant, first-order systems 513–14 translational springs 121–2 definition 17
time-delay parameter 44 translational systems multiplication operations 21–3
time-domain block diagram 457–64 derivation of equations of motion scalar multiplication 19
nonlinear system 11 130–8 unit-vector decomposition 19–21
Simulink models on 465–9 effective spring coefficients, elastic vector multiplication operations 21–3
time domain, system response analysis bodies 122–4 velocity-sensitive microphones 746
502–44 free-body diagrams and auxiliary vibration analysis of car moving 667
concepts 503–8 plots 127–9 dynamic modeling 669–73
first-order systems 508–14 friction forces 125–7 potholes, time response 681–6
higher-order systems 539–44 lumped masses 120–1 quasi-static deflections and dynamic
second-order systems 515–28 transfer function formulation 138–44 response 675–6
sinusoidal input, second-order translational springs 121–2 sinusoidal profile, steady-state
systems to 537–9 viscous dampers 123, 125 vibration 676–81
underdamped second-order systems transmissibility parameters 570–1 state-space representations
528–37 trans-resistance amplifier 296 673–5
time-invariant linear system 408–9 triangular pulse, piecewise-continuous system description 667–9
time-invariant models vs. time-variant function 61, 62 viscous dampers 123, 125
models 5 trigonometric manipulations, complex viscous friction 125–7
time-invariant systems 95 numbers 36–7 voltage divider rule 235–7
block diagrams 97–8 Tsiolkovsky rocket equation 137 in potentiometer 239–40
transfer functions 95–7 turbulent flow 361 voltage drop 238–41
time responses two-mass dynamic system 65–6 voltage follower 294
car moving over pothole 681–6 two-segment commutator 435–7 voltage-isolation amplifier 277–9
feedback control system 613, 615 typical inverting op-amp circuit 282 voltage-to-current converter circuit
first-order systems 508–14 295–6
higher-order systems 539–44 undamped system 517, 520, 522, 525,
second-order systems 515–28, 537–9 537–8 water hammer 368
time-shifted function 44–5 underdamped second-order systems 528 water purification (liquid-level system)
time-variant system 8, 136, 457, 471, 482 model parameters 529 711–12
torsional springs 147–9 performance specification actual outflow 713
traditional condenser microphone 754–5 parameters 529–35 description 712–13
trans-conductance amplifier 295–6 physical parameters 535–6 equilibrium point 719
transfer functions 6–7, 168, 399 sets of parameters 535–7 generated liquid-level graphs 715, 717
control systems, formulation 592–5 underdamped system 517, 520, 523, governing equations 713
derivation of 139 525 linearization 719
formulations 401, 408–13 unit conversion factors 760 Mathematica code 715, 716
negative feedback configuration unit matrix 24 nonlinear and linearized system, PI
416–17 unit ramp function 43 control 722, 725
quarter-car suspension model 142 unit step function 43 nonlinear state-space model 714–16
spring–mass–damper system 140–4 unit vector piecewise-continuous function, PI
state-space representations and 186, decomposition 19–21 control 722–3, 724
469–75 definition 17 state-space model 714
time-invariant linear system 408 unity-feedback control system 608 steady-state liquid levels 715, 717, 718
786 Index

water purification (cont.) dynamic system, modeling/ time response of car 683–6
system parameters variability 718 simulation 475–85 transfer function formulation
transfer function and PI control 721 equation solvers 776 776
Wheatstone bridge 724, 726, 727, graphics generation and styling
730 functions 774–5 zero matrix 24
Wolfram Mathematica 3, 99, 100, mathematical functions 773 zero-pole-gain form 41, 409
198–203, 297–306 matrix functions 774 zero vector 17
classical analysis and design operators and special characters 772 Ziegler, J. G. 606
777 response 777 Ziegler–Nichols ultimate-cycle method
drop-down menu 771, 772 state-space formulation 777 607–8, 638