Dynamic Model of a

Hydraulic Servo System for a

Manipulator Robot

Yalcin Efe

Degree project in
Electrical Machines and Drives

Stockholm, Sweden 2014

XR-EE-E2C 2014:002
 Master of Science Thesis

Dynamic model of a hydraulic servo system for a

manipulator robot

Yalcin Efe

Approved Examiner Supervisor

2014-01-24 Stefan Östlund Stefan Östlund, Fredrik Nilsson

Abstract
In this master thesis, a mathematical model of a hydraulic servo system for a manipulator robot
is completed by using several different methodologies. The models proposed are particularly
tuned for the DeLaval VMS robotic arm. The parameter identification of the robotic arm is
accomplished by dividing the model into several subsystems and investigating each system
separately by using catalogue data, experimental data and construction drawings. Furthermore,
the assumptions are proposed based on the literature review and the expertise of in-house
engineers.
After completion of parameter identification several different mathematical models including
linear and nonlinear methodologies are introduced. It is demonstrated that the improved-
nonlinear model can successfully mimic the movement of the robotic arm with relatively small
errors and it is found to be fairly reliable. Moreover, the errors incurred when chamber pressures
are compared with experimental data are found to be relatively small. Furthermore, the improved
linear model have successfully delivered an accurate position estimation especially for the
medium valve opening, while the chamber pressures are relatively less accurately predicted.
The study further carries out sensitivity (uncertainty) analyses to investigate the crucial
parameters of the model since it is sometimes very problematic to precisely estimate these
parameters. It is found out that the flow coefficient and supply pressures have remarkable impact
on the results of the simulations. Therefore, it is strongly advised that these parameters should be
very carefully evaluated during the modeling process.
Finally the bulk modulus models are compared and the influence of the bulk modulus is
revealed.

Page I
P a g e II
 FOREWORD

First and foremost, I owe my deepest gratitude to Fredrik Nilsson for guiding me through the
project with the right feedback, advice and support, and making me feel comfortable at DeLaval.
This thesis project would not have been possible without you.

In a second instance, I would like to thank to my supervisor Stefan Östlund and my coordinator
Nils Brandt for their guidance and feedback through the project.

I want to also express my gratitude to Ricardo Giscombe and all the colleagues in DeLaval, since
they were always ready to help and to give me the right working environment through the year.

Finally, yet importantly, words are inadequate in offering my thanks to my beloved family for
their love and blessing.

Yalçın Efe
Stockholm, November 2013

P a g e III
P a g e IV
 NOMENCLATURE

Notations
Symbol Description
Discharge coefficient (-)
σ Viscous friction parameter (-)
Density (kg/m3)
Time constant (s)
Loss factor (-)
α Piston area ratio (-)
Area (m2)
A Piston area (m2)
c Wave speed in oil (m/s)
Hydraulic capacitance of chamber A (m3/Pa)
Hydraulic capacitance of chamber B (m3/Pa)
Stribeck velocity parameter (-)
Flow coefficient (m3/sec √ )

. Compensated flow coefficient (m3/sec √ )

Valve spool diameter (m)
E Effective bulk modulus (Pa)
E’ Effective bulk modulus (Pa)
Force (N)
Valve overlap function (-)
Coulomb friction (-)
Coulomb friction parameter (-)
Fext External force (N)
Ff Friction force (N)
FL Load force (N)
Static friction (-)
Stribeck velocity parameter (-)
Viscous friction (N)

Page V
 , Valve hysteresis compensation parameter (-)
Valve input signal compensation parameter (-)
Capacitance term (m5/Pa)
Valve control signal term (1/sec)
, Flow-pressure coefficient at chamber A (m3/sec)
, Flow-pressure coefficient at chamber B (m3/sec)
Valve sensitivity coefficient
, Flow gain at chamber A (m3/sec )
, Flow gain at chamber B (m3/sec )
mA,fl Fluid mass at side A (kg)
mB,fl Fluid mass at side B (kg)
Piston mass (kg)
Total mass (kg)
Pressure (Pa)
pA Chamber pressure A (Pa)
pB Chamber pressure B (Pa)
Nominal pressure drop (Pa)
pS Supply pressure (Pa)
pT Tank pressure (Pa)
Fluid flow rate (m3/s)
Flow to chamber A (m3/s)
Flow to chamber B (m3/s)
External leakage flows (m3/s)
Internal leakage flow (m3/s)
Nominal flow (m3/s)
t Time (sec)
Damping term (Pa/sec)
* Valve input signal (-)
V Control volume (m3)
VA Piston chamber volume (m3)
VB Ring chamber volume (m3)
Vpl,A Pipeline volume at A side (m3)
Vpl,B Pipeline volume at B side (m3)
Piston position (m)

P a g e VI
 Spool position (-)
y(t) Hydraulic servo system output (m)
yref(t) Reference signal (m)
z(t) Disturbances (N)

Abbreviations
CCD Charge-coupled Device
HSS Hydraulic Servo System
LTI Linear Time-invariant
PI Proportional-integral
VMS Voluntary Milking System

P a g e VII
 TABLE OF CONTENTS

1 INTRODUCTION 2
1.1 Background 2
1.1.1 VMS multipurpose robotic arm 2
1.2 Purpose 3
1.3 Limitations 3
1.4 Outline 4

2 INTRODUCTION TO HYDRAULIC SYSTEMS 5

2.1 Hydraulic Servo Systems 5
2.1.1 Fundamentals of Hydraulic Servo Systems 5
2.2 Delaval’s hydraulic robot arm 10
2.3 Overview of the models proposed in the literature 11

3 MATHEMATICAL MODELING OF HYDRAULIC COMPONENTS 15

3.1 The fundamental equations of fluid mechanics 15
3.1.1 Bulk Modulus 15
3.1.2 Flow through orifices 18
3.1.3 Flow equations in valves 18
3.1.4 Continuity equation and pressure transients 19
3.2 Elementary models of HSS components 20
3.2.1 Valves 20
3.2.2 Pipelines 25
3.2.3 Hydraulic cylinder 26

4 SIMULATION MODELS AND MEASUREMENT ACQUISITION 30

4.1 Characterization and interconnection of subsystems 30
4.2 Structured nonlinear model 31
4.2.1 Assumptions 31
4.2.2 The model description 32
4.3 Structured linearized model 33
4.3.1 Assumptions 34
4.3.2 Model description 35
4.3.3 Parameter feedback 35

P a g e IX
4.4 Simplified linearized model 36
4.4.1 Parameter feedback 37
4.5 Further information on essential model parameters 37
4.5.1 The valve flow‐signal function 38
4.5.2 Valve overlap and underlap 38
4.5.3 Actuator dimensions and mass 38
4.5.4 Friction forces 39
4.5.5 Leakage coefficients 39
4.6 Simulation package 40

5 RESULTS 41
5.1 The introduction of measured data 41
5.1.1 Friction forces 42
5.1.2 The flow coefficient 42
5.1.3 The valve hysteresis 43
5.2 Comparison of models 44
5.3 Comparison of improved models 50
5.4 Comparison of all proposed models 50
5.5 Sensitivity Analysis 57
5.5.1 Influence of flow coefficient 57
5.5.2 Influence of valve response time 57
5.5.3 Influence of supply pressures 58
5.5.4 Influence of mass variation of the system 58
5.5.5 Influence of possible external forces 58
5.6 Further concerns on effective empirical bulk modulus 58

6 DISCUSSION AND CONCLUSIONS 65

6.1 Discussion 65
6.2 Conclusions 66
6.3 Future work 67

7 REFERENCES 68

Page X
 1 Introduction
This chapter provides a brief introduction to the Delaval Voluntary Milking System and sets a
ground for the purpose and the limitations of the study. Finally, the chapter ends with a short
description of the project outline.

1.1 Background
DeLaval has around 125 years of innovation and experience in the dairy industry by supporting
dairy farmers to manage their farms in an appropriate way (DeLaval, 2013). The Voluntary
Milking System (VMS™) is a milking system that is developed for automatically milking cows.
It enables cows to decide if they want to be milked and the system harnesses up to 3000 liters of
milk per day per unit.

1.1.1 VMS multipurpose robotic arm

DeLaval VMS robotic arm is an advanced robotic arm consists of two links driven by three
hydraulic actuators. The dynamic control of the arm is maintained by a hydraulic system and;
hence, it is relatively fast, quiet and robust. The arm can adjust itself to various type of cows and
this leads to longer lifetime milk production per cow. Moreover, the hydraulic power provision
benefits lower energy consumption and lower maintenance costs (DeLaval, 2013).
The main purpose of the arm is to handle the teat preparation before milking (including optional
pre-spray), to attach and reattach the teat cups, to align the milk tubes and lastly to take-off the
cups and spray the teats.
When the cow enters the VMS, firstly the gripper fetches the teat cleaning cup from the teat cup
magazine and sterilizes the teats. Afterwards, the teat cups are attached to the teats by a gripper.
It should be noticed that the fast and accurate teat localization for a quick and high attachment
rates are sustained by an optical CCD camera equipped with dual lasers given that the position
of the teats differ for every cow. Once the milking is completed, the teats are disinfected by
spraying a chemical shim from a mouthpiece on the arm.

Page 2
 Figure 1.1. Delaval VMS multipurpose robotic arm (DeLaval, 2013)

The arm is modeled according to a human arm’s flexible movement range. The general benefits
of the advanced robotic arm can be listed as:

 low energy consumption

 high safety for humans and cows due to low running pressures
 fast and accurate dynamics
 requirement for very little maintenance
 compact design.

1.2 Purpose
The main purpose of this thesis is to obtain a flexible and robust model that can simulate
hydraulic servo dynamics of the robot arm integrated in DeLaval’s foremost voluntary milking
system. In this context, the main goals are

 to simulate the hydraulic dynamics of the robot arm and to validate the model by using
experimental setup built by DeLaval
 to investigate the crucial parameters and reveal the influence of these parameters on the
model reliability.

1.3 Limitations
The thesis is carried out in cooperation with DeLaval; therefore, the model intensively focuses
on the current robot arm used in the VMS. The other limitations of the study are:

 Neither pipeline dynamics between the valve and the actuator nor between the supply
and the valve are included.

Page 3
  This study concentrates on the hydraulic dynamics of the robot arm rather than
investigating the entire dynamics of VMS robotic system.
 The system is investigated in open loop manner and a development of a complicated
controller is avoided.
 A Quasi-two-dimensional1 experimental setup is built for the validation of the simulation
model; however, the influence of the gravity is further discussed.
 The load force on the actuator and flow through valves are not directly measured from
the experimental setup. These quantities are calculated indirectly.

1.4 Outline
This study begins with an introduction to hydraulic systems. A literature review is included in
the introduction. Chapter 3 gives a detailed description of mathematical models for hydraulic
servo systems. Then, in Chapter 4 the description of models that have been used for the
simulation of Delaval’s hydraulic robot arm and further assumptions made in the process of
building the simulations are discussed. Chapter 5 presents the results that are obtained with the
process/methods described in the previous chapters and the final chapter is devoted to
discussion of the results and recommendations for future work.

1
The experiments are intented to be implement only x and y axis. However, small deviations occur in the z axis and
this phonemena requires further considerations on the gravity. Further discussions are introduced in section 5.1

Page 4
 2 Introduction to Hydraulic
Servo Systems
This chapter is intended to deliver an introduction to hydraulic systems and present a brief
overview of the relevant aspects of the area. The identification of Delaval’s hydraulic system is
also presented in this section. The chapter ends with a detailed literature review.

2.1 Hydraulic Servo Systems

With the beginning of 20th century the fluid power technology rapidly evolved with the help of
the first generation hydraulic drives that basically compromise some flow control devices that
are driving the hydraulic actuator in an open loop manner (Jelali & Kroll, 2004).
After the Second World War, the servo control industry had built up sufficient knowledge to
implement accurate closed loop techniques. After that day, the closed loop servo control
systems had rapidly advanced and this introduced a wide range of applications where these
systems can be practiced.
In recent years, electrical drives have become more popular in the domain of high-performance
motion control. However, hydraulic servo-systems are still present in wide diversity of
applications that include current industrial motion systems such as machining plants, robotics,
motion simulators, fatigue testing systems and so on (Walters, Eng, & Mech, 1991).
This chapter presents an overall description of hydraulic servo systems with a brief summary of
several subsystems that are often included in a typical hydraulic servo system.
2.1.1 Fundamentals of Hydraulic Servo Systems
It is possible to divide a hydraulic servo system into a few basic groups of individual
components, interconnected to perform a specific function in the system. In this context, a
simple structure of a hydraulic system mainly comprises
 hydraulic power supply
 control elements
 actuators
 other auxiliary elements (data transmission elements etc.) (Walters, Eng, & Mech, 1991)

Page 5
The hydraulic power supply generally consists of a pump and ancillary equipment such as
accumulators, relief valves etc. The pump transforms the mechanical power from a prime mover
(electric or diesel motor) into hydraulic power at the actuator. Furthermore, valves are used to
regulate flow direction, power, and the amount of fluid and pressure delivered to the actuator.
Finally, an actuator (e.g. linear or rotary) converts the hydraulic power into the desired form of
mechanical power at the output.
The fluid medium lubricates the components, seals in valves and cools the system while
providing direct transmission and control of the power.
Various system components are linked together with connectors to regulate the power of the
fluid under pressure.
Finally, the fluid storage and conditioning components ensure necessary quality and quantity,
and cooling of the fluid.

2.1.1.1 Hydraulic Power Supply

The source of a hydraulic power in servo systems can generally be regarded as pumps.
Hydraulic servo-systems often practice pressure controllers in conjunction with variable-
displacement pumps, but other controllers may be used to regulate flow, pressure and flow
combined, or input power (Younkin, 2003).
There are different configurations in which the power supply circuits can be designed. In an
open circuit system, the pump draws the fluid from a tank, which is utilized by the system, and
the return flow exhausts back to the tank. In a closed circuit, there is a continuous flow between
a motor and a pump. This effectively means that the pump output flow is sent directly to the
hydraulic motor and then returned back in a continuous motion to a pump.

2.1.1.2 Control elements

Hydraulic control valves are devices that use mechanical motion to control a source of fluid
power. Thus, it is fair to say that the valves are the most crucial mechanical (or electrical) link to
the fluid interface in hydraulic systems. Based on their functional purposes, they diverge in
arrangement and complexity (Merritt, 1967).
There are principally four different classes of valves in hydraulics (Jelali & Kroll, 2004):

a) Pressure valves are commonly used to maintain the system in a predetermined pressure
level to ensure the system safety
b) Check valves are a special class of directional valves since they only allow fluid flow in
one direction while blocking the flow in the reverse direction.
c) Flow control valves regulate the flow rate of a fluid. For example, they may be used to
limit the maximum speed of cylinders or motors.
d) Directional control valves are one of the most essential devices in hydraulic systems.
They permit the fluid to flow into various paths from one or more sources. They

Page 6
 generally contain a spool inside a cylinder that is mechanically or electrically regulated.
The position of the spool restricts or permits the flow; thus, it controls the fluid flow.
Therefore, the proportional variable control enables infinitely adjustable and rapid
variable settings of actuators based on force, speed and stroke position.

The configuration of directional control valves can be classified according to (Merritt, 1967):

a) the number of “ways” that flow can enter and leave the valve. Since all the valves
require a supply and a return; and often at least one line to the load, valves are frequently
equipped either three-way or four-way.
b) the number of switching (or discrete) positions. This number varies from one in a
primitive valve to the usual two or three. However, up to five positions or more can be
practiced with special valves.
c) the technique of valve actuation that forces the valve mechanism to move itself to an
alternative position. Either external signal commands such as electrical, manual or pilot
practice or internal signal commands such as pilot pressure or spring force could be
realized to move the valve mechanism.

2.1.1.2.1 Centre types of the directional control valves

Type of the valve center may directly affect the certain characteristics of a valve. In this sense, a
“lap” represents the width of the lands relative to the width of the ports in the valve bore (see
Figure 2.1.a).
An open center or underlap valve’s land width is smaller than the port width in the valve sleeve.
This effectively means that for a small amount of time all the valve ports could be connected to
each other. In return, this leads to a smooth, pressure-peak-free switching during the cross-over.
On the other hand, there is a possibility that undesired actuator movements may occur and; thus,
they are often preferred in closed-loop applications (see Figure 2.1.b) (Jelali & Kroll, 2004).
A closed center or overlap valve’s land width is larger than the port in the valve sleeve. In this
context, for a small amount of time all the valve ports are disconnected from each other. This
prevents system pressure on the actuator to collapse during the cross-over. However, this can
also lead to undesirable pressure peaks (Totten, 2000).
A critical center or zero valves’ land width is equal to the port in the valve sleeve. The majority
of commercial valves are listed as zero-lapped valves, basically they can obtain linear flow-
signal curve (Jelali & Kroll, 2004).

Page 7
Figure 2.1. a) Definition of center types b) their corresponding flow signal graphs c) leakage flow curves (Jelali &
Kroll, 2004)

2.1.1.3 Actuators
A hydraulic actuator converts hydraulic power provided by a pump and processed by control
elements into useful mechanical work. The actuators can produce linear, rotary, or oscillatory
output.
One of the most frequently used hydraulic actuator is a hydraulic cylinder (also called a linear
hydraulic motor) that is used to generate a unidirectional force through a unidirectional stroke.
Single acting actuators allow the application of hydraulic power only in one direction while
double acting actuators enable users to practice hydraulic force in both direction (Ilango &
Soundararajan, 2012).
Double-acting actuators with rods on both sides are called symmetric/synchronizing cylinders
and can provide equal force generation in both directions. It should be also noticed that the void
volume filled by the hydraulic fluid is also identical for these hydraulic cylinders; hence, the
subsequent piston speeds are also the same. However, the symmetric cylinders are greater in
length and consequently more expensive than double acting cylinders. Therefore, the majority of

Page 8
actuators that are used in the industry are as double acting cylinders (Jelali & Kroll, 2004) (see
Figure 2.2).

Figure 2.2. A simple representation of a double acting cylinder

2.1.1.4 Other elements

Control loops: Generally, one or more hydraulic actuator outputs are used for controlling
purposes; and, this forms a closed loop hydraulic control system as shown in Figure 2.3.

Figure 2.3. Block diagram of a hydraulic servo system in a closed control loop

The electronic measuring devices sense the system output y(t) and based on this measurement,
the output is forced to follow the reference signal yref(t) even though there are disturbances due
to load; z(t) and due to sensor errors; v(t) (see Figure 2.3). If a deviation from the reference
signal occurs (means an error at summing point), the controller undergoes a corrective action
since the error is fed back to the controller until it vanishes. In this context, it is crucial that the
tracking is accurate, even though the dynamics of the plant could change during operation.
On the contrary to a closed-loop system, an open loop system has no feedback from the load.
Thus, correction is not possible if the load behavior deviates from the desired projection.
However, it should be noted that there exist many applications where an open loop control

Page 9
system is adequate enough and able to avoid the extra cost of a closed loop system (Younkin,
2003).
Sensors/Transducers: Many of the current servo system applications require a high degree of
accuracy in the system response and hence require precise devices to measure the controlled
variables or other state variables such as (Jelali & Kroll, 2004):

 pressure
 flow
 position (angle)
 velocity (speed)
 acceleration/deceleration
 flow.

It should be also understood that the accuracy of measured parameters is crucial since the
controlled variables can never be more precise than the measuring device’s tolerances. This also
implies that the frequency, at which the measurements are taken and transferred to the electronic
controller, is essential since it can directly affect the system’s response and dynamic behavior
(Norvelle, 1999).

2.2 Delaval’s hydraulic robot arm

Among the wide variety of hydraulic actuator and valve combinations, Delaval’s robotic arm is
based on the valve-cylinder combination shown in Figure 2.4. The system consists of a
proportional valve and a differential cylinder (sometimes also referred as single rod), identified
as A+αA configuration and most a commonly used valve-actuator combination due to its
compact size (Jelali & Kroll, 2004).
A control signal u is applied to the valve as a voltage. The main purpose of the input signal is to
regulate oil flow through the valve orifices. The power supply unit maintains variable oil supply
pressure pS, while the return flow fed to a tank is at around atmospheric pressure pT.
The piston divides actuator volume into two different chambers, given as chamber A and B.
Fluid flows in or out off the chambers resulting in the movement of the piston. This fluid flow
varies the pressures in the chambers, pA and pB respectively, and generates the required pressure
difference across the chambers to mobilize the load at the actuator. In this sense, it can be
noticed that the piston motion depends on the load of the actuator.
A constant oil flow is maintained by a constant valve opening and pressure drop over the valve
which in turn results in a linear movement of the piston. However, it should be stressed that the
oil is compressible; thus, the chambers act as two springs.

P a g e 10
2.3 Overview of the models proposed in the literature
The literature survey is started with an investigation of general strategies on modeling and
simulation. In this context, a very enlightening study is carried out by Maria (1997) to provide
introductory tutorial on modeling, simulation and analysis. The main questions in the study are
“What is modeling and simulation? What is simulation modeling and analysis? What types of
problems are suitable for simulation? How to select simulation software? What are the benefits
and pitfalls in modeling and simulation?”. The author claims that the intended audience is those
who are involved in system design and modification - system analysts, and the study delivers a
road map for developing a simulation model, designing a simulation experiment, and
performing a simulation analysis.
In this study, once the road map is drawn and the answers of above questions are found; more
specific literature analysis on hydraulic valve/cylinder systems is carried out. It is found out that
there are several different models that focus on the different aspects of the hydraulic system.
Menshawy, Moghazy, & Lotfy (2009) investigate another relevant case that explores the
dynamic performance of an electrohydraulic system, containing a proportional directional valve.
They build a model based on a verified theoretical model of the directional proportional valve,
which was previously presented and published by Menshawy (2006). They also verify their
simulation by constructing an experimental setup and cross-checking the results. It should be
noted that the simulation is built in Matlab/SIMULINK environment.
Tenali (2008) studies a larger scale of electro-hydraulic servo actuator and builds also a
mathematical model for applications like aircraft flight control machinery. The author uses
Matlab/Simulink models of the actuator and its components and the time response of the linear
actuator is also derived by using Matlab/Simulink Software.
It should be noted that even though these studies have different settings, they all are built in
Matlab/SIMULINK. The authors given above agree that the available simulation blocks in
SIMULINK library reduce the time required for implementation of such models. Moreover, it is
also easy to implement a hydraulic system in SIMULINK since there are turn-key block models
for some hydraulic components such as valves and actuators.
Dasgupta & Murrenhoff (2011) prefer to examine their closed-loop servo-valve controlled
hydro-motor drive system by using Bondgraph simulation technique. They validate their
findings of servo-valve model by a comparison with the established results of Gordic et al.,
(2004). The further focus is established on the variation of the parameters of the PI controller
that regulates the servo-valve in their study. A very similar study is also carried out by Dasgupta
& Watton (2005); however, they focus on the different pressure-flow characteristics across the
valve ports and the orifice.

P a g e 11
Figure 2.4. Valve-cylinder combination with power supply

P a g e 12
One of the most interesting research is carried out by Eryilmaz & Wilson (2006). The authors
develop nonlinear equations for a generic proportional valve model and these equations are
used to obtain simplified flow rate expressions under commonly recognized assumptions.
These equations rely on a set of geometric spool properties and physical model parameters for
the flow rate through the valve ports. The study focuses on procurement of a single set of flow
rate equations that are applicable regardless of valve center type. They argue that these unified
model equations are useful for simulations and nonlinear controller designs.
At this point, it should be pointed out that contrary to SIMULINK models, Matlab script
models are often more flexible since users can adapt any part of the script to any other script
softwares without a requirement of SIMULINK simulation packages. Moreover, the script
models often provide more transparency for its user. Besides, the literature study reveals that
the authors are mainly using commonly agreed flow and pressure equations with minor
changes. These modifications are mostly done to adapt general equations to more specific
cases with reasonable assumptions.
There are also more component or parameter focused studies present in the literature. For
example, Mitianiec & Bac (2011) have developed a mathematical model for the hydraulic
valve timing system. It focuses on the complex physical phenomena that occur between the
valve components and the hydraulic liquid. The mechanical parts of the valve are simulated by
SIMULINK while the electric control parameters in order to acquire required lift and timing of
valves are obtained with the help of the YARIS SI 1.3 l engine. Similarly, Wei & Xianxiang
(2010) propose a second order differential equation to describe the relationship between the
effective orifice area and the input signal. On the contrary, Kilic, et al., (2012) develop a model
to predict accurately the pressures of a servo-valve controlled hydraulic cylinder accurately by
using advanced modeling tools such as artificial neural networks.
The final remark can be made on the book of Jelali & Kroll (2004) since it presents a complete
description for each part of a hydraulic-servo system with many different implementation
options. However, it should definitely be noted that some of the equations in the book have
typos or calculation mistakes; thus, it is advised to readers that not all the equations in this
book should be taken as granted.
In this context, it can be concluded that the model and the simulation used in this study is built
on using several different modules/concepts proposed in the literature. It is fair to say that the
most of the simulation is built upon the suggestions by Jelali & Kroll (2004). However,
Eryilmaz & Wilson (2006)’s unified equations are found to be very practical in several
occasions and implemented in the current model. Moreover, the SIMULINK models are used
as a guide to generate relevant scripts and many modules are translated into Matlab script in
the light of previously proposed models. Finally, further improvements on specific components
such as valves and hydraulic actuator are made based on the advices given by Wei &
Xianxiang (2010) and Kilic, et al., (2012).

P a g e 13
The executed model can be regarded as a specific implementation of models proposed in the
literature and an experimental setup built by DeLaval. Further improvements are made on the
external forces as a result of the movement of the hydraulic robot arm.

P a g e 14
 3 Mathematical modeling of
hydraulic components
This chapter describes mathematical models of essential hydraulic components and comprises
the most relevant dynamic and non-linear effects encountered in hydraulic servo-systems.
Necessary assumptions are made to keep the model as simple and relevant as possible. It must
be noted that the complete theoretical discussion of fluid motion in fluid mechanics is not the
scope of this study.

3.1 The fundamental equations of fluid mechanics

Fluids are described as bodies that continuously and permanently deformed under shear stress.
It is also proven that the liquids cannot return to their original state after the deformation (Bar–
Meir, 2011). In this regard, the normal tension on the surface component of a fluid is given as
pressure. The magnitude of the pressure at the given point is identical in all directions. The
unit is given as force per unit area in a static fluid and can be calculated as:

3. 1

3.1.1 Bulk Modulus

It is commonly assumed that the hydraulic fluids are incompressible. However, the fact
remains: the fluids present some degree of compressibility (Totten, 2000).
The assumption of an incompressible fluid can be very practical in systems where a tight
control of the response is not required and the operation pressures are fairly moderate. In other
cases where high pressures(e.g. larger than 100 bars) are applied to a large volume of fluid; a
noteworthy amount of energy may be used to compress the fluid by squeezing the fluid’s
molecules closer together. This leads to a delayed response of the system meaning that a
loaded actuator may not move until upstream fluid has been compressed enough. Besides, a
compressed fluid stores the energy and this stored energy may cause the actuator to continue its
movement even after the closure of the control valve has closed (Jelali & Kroll, 2004).

P a g e 15
In this context, George & Barber (2007) define the bulk modulus briefly as:

“The bulk modulus is a measure of resistance to compressibility of a fluid“

In a hydraulic system that operates in the high pressure range, the bulk modulus should be
carefully considered since ignoring the bulk modulus could compromise the response time of
the system. Besides, applied pressure levels can affect the behavior of the system rather than
only compressing the fluid. Therefore, it is essential to design systems with as little fluid as
possible between the valve and the actuator (Cundiff, 2002).

3.1.1.1 Influence of entrained air

It is common experience that the hydraulic fluids are aerated in use. Therefore, the entrained
air in the hydraulic fluids can lead to significant variations on bulk modules since the air is
much more compressible than the oil (George & Barber, 2007).
It should be stated that there are several equations presented in the literature for the estimation
of the bulk modulus such as by Jelali & Kroll (2004), Totten (2000) and Wang, Gong, & Yang
(2008); however, the expressions given in these proceedings require an accurate determination
of many quantities (e.g. the volume of the air entrained in liquid), and thus the experimental
data are difficult to use in practice. On the other hand, the agreement made on that the
influence of entrained air is generally more significant in the low-pressure region. This
phenomenon occurs due to the solubility of air in fluids and increases with pressure. The
designers should also be aware of that the air dissolved in a fluid at high pressure can form
bubbles in case of large pressure drops — a phenomenon which is known as cavitation.

3.1.1.2 Influence of temperature

Another important parameter is the fluid temperature since fluids tend to expand with
increasing temperature, which in turn leads to increase in pressure (Cundiff, 2002).
It should be pointed out that compressing a fluid leads to an increase at the fluid temperature.
In a rapid compression, the generated heat cannot be dissipated quickly and this may cause
even further pressure increase. However, the generated heat can be dissipated when
compression occurs slowly. Thus, the system behavior must be carefully investigated to avoid
such phenomenon.

3.1.1.3 Effective bulk modulus

It should be stressed that estimating all these variety of parameters could be very difficult in
practice. Thus, many researchers have driven empirical formulas for the estimation of effective
bulk modulus E’, which includes the effects of temperature and entrained air merged from
direct measurements. A frequently used estimation of effective bulk modulus for hydraulic
cylinders is given by Lee & Lee (1990)

P a g e 16
 ′ 3. 2

with the parameters 0.5, 90, 3, 18000 bar and 280 bar.
Eggerth (1980) expresses effective bulk modulus as

1
3. 3
/

with parameters and found in Table 3.1; is assumed to be 10 bar.

Table 3.1. Parameters of Eggerth’s formula (Eggerth, 1980)

Temperature (oC) (10-10 m2/N) (10-10 m2/N)

20 4.943 1.9540 1.480
50 5.469 3.2785 1.258
90 5.762 4.7750 1.100

The influence of the temperature on the effective bulk modulus, estimated by using Eggerth’s
formula is shown in Figure 3.1.

Figure 3.1. The influence of temperature on effective bulk modulus estimated by using Eggerth’s formula

Jelali & Kroll (2004) argue that these approximations are sufficient for design purposes;
although, experimental data are always preferable.

P a g e 17
3.1.2 Flow through orifices
The simple description of an orifice can be given as “an opening in a vessel, through which the
liquid flows out”. Orifices are sudden sharp edged restrictions in the flow passages. The
opening area of an orifice can be fixed and variable (Kundu, Cohen, & Dowling, 2011).
The purpose of orifices is to control the flow or to generate a pressure difference between the
chambers. Most of the orifice flows occur in high Reynolds numbers and this region is
assumed to be of major importance. The Reynolds number (Re) is a dimensionless quantity
which is defined to help prediction of similar flow patterns in various fluid flow conditions.
Flows with high Reynolds Number are regarded as turbulent flows (Jelali & Kroll, 2004).
The flow estimation in an orifice is governed by speed of the flow and the opening area of the
orifice and

2
3. 4

where is the density of fluid and is assumed to be the loss factor that depends on the
geometry and Reynolds number. Instead of using equation (3.4), it is more practical to use a
modified orifice equation given as

2
∆ 3. 5

The parameter is called the discharge coefficient and can be calculated theoretically by

/ 2 0.611 3. 6

This value can be used for the sharp edged orifices as long as the flow is turbulent (Jelali &
Kroll, 2004).

3.1.3 Flow equations in valves

Similar to orifices, the flow through valves are also defined by orifice equations with a linear
relationship between the valve spool position and the flow area given as

,∆ ∆ 3. 7

with the flow coefficient

P a g e 18
 2
3. 8

where is the diameter of the valve spool. In the equation above, the spool opening area is
.
It should be pointed out that the flow coefficient can be best estimated by an experiment.
However, such an experiment is generally difficult and requires very precise instruments (Jelali
& Kroll, 2004). Therefore, the flow coefficient is generally determined by using catalogue data
of the valve suppliers with the help of the formula:

1
3. 9
∆ ,

where is the nominal flow, ∆ is the nominal pressure drop, and , is the maximum
stroke of the valve.

3.1.4 Continuity equation and pressure transients

For a defined quantity of matter, control mass or control volume, it is possible to derive the
conservation laws for the mass, momentum and energy. For example, the integral form of the
continuity equation (mass) for the given control area seen in Figure 3.2 can be expressed as

0 3. 10

where the density , is often not constant in hydraulic servo systems (see section
3.1.1). This equation can be formulated in the differential coordinate-free form

A2
A1

ds
dV=Ads ds
(2)
(1)
Figure 3.2. The description of a control tube

P a g e 19
 0 3. 11

If the characteristics of the equation above are investigated, it would be seen that there exist
two special cases which are

0 3. 12

0 3. 13

Now it is assumed that the control volume is V and the stored fluid inside the control volume
has the mass of m with a mass density of ρ. Moreover, the flow rate is steady which implies
that the fluid mass stored inside the volume must be equal to the outgoing mass flow rate
subtracted from the incoming mass flow rate. Hence, the equation can be rewritten as

3. 14

In a constant temperature, the density change can be written in a linearized form that is given
as

3. 15

If (3.14) and (3.15) are combined

3. 16

Now, if the volume is fixed (V=V0), (3.16) becomes

3. 17

3.2 Elementary models of HSS components

The purpose of this section is to gather some of the most essential component models of
hydraulic servo-systems, including valves, cylinders, pipelines and power supplies.

3.2.1 Valves
The valves are rather complicated devices and Merritt (1967), Jelali & Kroll (2004) and Ilango
& Soundararajan (2012) argue that many non-linear dynamics are present. This, in turn, results
in a complex non-linear model. The non-linear effects most often encountered are

P a g e 20
  Dead band of the valves with overlap
 Square-root behavior of the flow function
 Complicated flow-induced forces in addition with friction forces
 Valve hysteresis i.e., the identical change in the controller output in both directions
leads to a different change in the process value.
 Saturations (velocity, flow, stroke etc.)
 Response sensitivity (resolution), i.e., the amount of input signal change required to
acquire a measurable valve flow change
 Repeatability, i.e., variation in output flow or pressure that may exist when the input
command is repeated

3.2.1.1 Pressure-flow equations for valves

Flow through a valve orifice is described by the orifice equation (3.5), which takes the
direction of the pressure drop (flow direction) into account. i.e.,

,∆ ∆ ∆ 3. 18

Even though (3.18) presents a simple valve model, it does not always hold in all valve
applications. More specifically, the spool might not be critically centered (zero overlap),
intentionally or due to manufacturing tolerances. For example, both supply and return orifices
may be closed at a given time for a closed center valve. Therefore, these concerns impose a
more complete valve model that can represent a wider variety of realistic operating conditions.

Figure 3.3. Overlapped 4/3 (four ports/three switching position) spool valve

In this context, flow relations for a proportional overlapped valve given in Figure 3.3 can be
rewritten as

| |
| | 3. 19

P a g e 21
 | |
| | 3. 20

where is the valve overlap and the function is defined by

0
3. 21
0 0

It should be noticed that the flow direction is determined by sign of and . The model
assumes that all the orifices are identical. In case 0, the equations hold valid for critically
centered valve.
The equations above also take into account a special case where the pressure of any chamber
exceeds the system pressure, e.g., in case of a sudden and rapid reversing of an actuator with a
large load. This effectively means that the cylinder would act as a pump in such a case.
Radial clearance and Leakage Flows: Equations (3.18-3.21) take into consideration the
pressure-flow behavior of valves. The valves often have radial clearance due to manufacturing
tolerances. Therefore, there always exist some extra leakage flows in valves which influences
performance, behavior and associated pressure-flow curves for small openings. Some other
complex models also consider the fact that rounded edges and radial clearances cause laminar
flow for small port openings (Merritt, 1967), (Jelali & Kroll, 2004) and (Mitianiec & Bac,
2011).
However, it should be stressed that outside these regions, the equations given above fit very
well. Besides, these non-linear dynamics can be added to the model by changing cv to

. 3. 22

The compensation parameter varies between 0 and 1 based on the valve input signal.
The value of can be determined by using catalogue data; however, a more accurate
description can also be made by using experimental data in case it is possible.
Valve hysteresis: Hysteresis is the point of widest separation between valve flow/input signal
characteristic curve with increasing input relative to decreasing input. This phonemena can be
seen in Figure 3.4 for a proportional valve with an overlap. This effect can be compensated by
an additional parameter , that depends on the valve input signal and flow direction

. , 3. 23

for the valve hysteresis and varies motsly between 0.80 and 1.

P a g e 22
 Figure 3.4. The valve hysterisis (Johnson, 2013)

Valve dynamics: The valve dynamics involves a large set of parameters that are seldom
known within some range, and mostly completely unknown. Information on these parameters
is generally acquired from diverse sources such as manufacturer’s catalogues and drawings,
literature and heuristic/manual optimization. On the other hand, an accurate analytical
description would be time consuming and extremely difficult to verify (Dasgupta & Watton,
2005), (Gordic, Babic, & Jovicic, 2004), (Jelali & Kroll, 2004) and (Merritt, 1967).
Nowadays manufacturer catalogues provide step responses and time constants and this
information can be used to model simplified valve dynamics such as a first order system
(Fisher, 2013).
The response time of the valve is defined by a parameter called T63, which is the duration
measured from the beginning of the input signal change until the output reaches 63% of the
corresponding transformation (Fisher, 2013). Here, the valve dynamics approximate a first
order system with a time constant of . First order linear time-invariant (LTI) systems are
characterized by the differential equation

1
3. 24

where describes the exponential decay constant and V is a function of time t

3. 25

The right-hand side of (3.25) is defined as the forcing function f(t). It describes an external
driving function of time, which can be given as the system input, to which V(t) is the response,

P a g e 23
or system output. A very classical example of f(t) is the Heaviside step function, denoted by
u(t)

0, 0
3. 26
1, 0

In this context, if the forcing function is replaced with a step input, the equation becomes

1
3. 27

The general solution to (3.27) for time t > 0, given that V(t=0) = V0 can be formulated as

/ /
1 3. 28

It should be noted that the long-time solution would be independent of time and the initial
conditions.

3. 29

The step response of a system for initial value V0=0, A=1 and 30 can be seen in Figure
3.5. It can be easily observed that the output is estimated as 63% of the input after 30 is
reached.

Figure 3.5. The step response of the valve for 30

P a g e 24
3.2.2 Pipelines
The connection of hydraulic components is carried out by pipelines. Jelali & Kroll (2004)
argue that the whole pipeline volume can be included in the corresponding cylinder chamber as
an inefficient volume if the length of the pipeline is smaller than a certain limit. This limit is
given by

/10 3. 30

Ff
Fc0 + Fs0

Fc0
Friction force

Fs
Fv
0

0
Velocity
Figure 3.6. Velocity-dependent friction force (Stribeck curve)

where c is the wave speed in oil and is the maximum value of the interesting frequency. If
the estimated length exceeds the length of pipeline, an additional simulation model for the
pipelines must be created.
Goodson & Leonard (1972) suggest using (3.31) for estimation of wave speeds in rigid lines

3. 31

The bulk modulus in the (3.31) can be replaced by the effective bulk modulus E’ and in this
way the influence of entrained air and mechanical compliance can be taken into account (see
section 3.1.1).
Applications when pipelines have to be modeled are listed below:

P a g e 25
 1. In the application where the oil supply unit is detached from the hydraulic servo system
with long connecting lines (Jelali & Kroll, 2004). This type of configuration can lead to
undesired oscillations: it should be appointed that Viersma (1980) has investigated
these systems and provided a lot of insight about the dynamics of such systems.
2. In some applications the servo valves cannot be located on the actuator chambers due to
space restrictions. Such a setup results in a relatively long pipeline between the valve
and actuator chambers, hence the pipeline modeling is required in these cases (Jelali &
Kroll, 2004).
3. In some very specific cases, the modeling of supply and return lines stands in the heart
of the modeling process (e.g., aircraft brake hydraulic systems). However, these
pipelines include quite different parts, such as rubber hoses etc. This, in turn, may have
significant impact on the generation of considerable pressure peakings and oscillations;
hence, an additional hydraulic notice.

It should be stated that further investigation of hydraulic pipelines and some specific models
are given in the literature by Jelali & Kroll (2004) Novak, Guinot, Jeffrey, & Reeve (2010) and
Younkin (2003) in case the reader desires to know more about pipeline modeling. However,
these more sophisticated models are out of the scope of this research simply because of the
given conditions above: they do not apply to the model that is investigated in next chapter.

3.2.3 Hydraulic cylinder

The theoretical modeling of hydraulic cylinders is well-known and the basic principles are
given earlier by various authors such as Viersma (1980) and Merritt (1967). However, an
experimental setup can improve the accuracy of the parameters used in the cylinder model.
Therefore, the theoretical model is slightly simplified to develop a compact model that can be
easily identified by experiments.
The most important non-linear effects that are essential for the cylinder model are

 Geometrical asymmetry due to the difference on the area of each side of the piston
 Pressure dependence of bulk modulus in combination with the fluid elasticity as well as
the elasticity of mechanical compliance
 “Actuator stiffness” dependent on the position i.e., the damping ratio of the transient
dynamics varies with cylinder position
 Friction forces involve with a highly non-linear Coulomb friction term.

It should be stressed that the compressibility of the oil influences the dynamics of hydraulic
servo systems remarkably. The hydraulic fluid shows a spring like properties and this
introduces a second order mass-spring system whose natural frequency restricts the bandwidth
of any hydraulic servo-system abruptly. The system can be damped by leakage flow and
viscous friction but their capabilities of damping are relatively small (Jelali & Kroll, 2004).

P a g e 26
3.2.3.1 Pressure dynamics in cylinder chambers
To begin with, (3.17) can be applied to each of the cylinder chambers, it yields

3. 32

3. 33

where VA is the piston chamber volume and VB is the ring volume chamber. The given volumes
include both the valve connecting line and the chamber volumes. Moreover, and
describe the internal and the external leakage flows respectively. The volumes of the chambers
can be given as

, 3. 34
2

, 3. 35
2

where Vpl,A and Vpl,B are the volumes of the pipelines at each side respectively. Therefore each
volume has an efficient part (i.e.,the volume of the chambers) and inefficient part (i.e., the
volumes of the pipelines between the valve and the actuator). The time derivatives of the
volumes can be formulated as

3. 36

3. 37

(3.32) and (3.33) can be reformulated to provide pressure dynamics in cylinders

1
3. 38

1
3. 39

The hydraulic capacitance of each chamber can be written as

,
, 3. 40

P a g e 27
 ,
, 3. 41

Internal leakage from one chamber to another can be given as

3. 42

where CLi is the internal leakage flow coefficient. The external leakage from chambers to case
drain can often be neglected.

3.2.3.2 The piston motion equations

The governing equation for the piston motion can be found by using Newton’s second law on
the piston forces. Thus, it becomes

∝ 3. 43

It should be pointed out that the total mass mt is sum of piston mp and the mass of hydraulic
fluid in the cylinder chamber and in the pipelines; mA,fl and mB,fl respectively:

, , 3. 44

Hence, the mass of the fluid can be calculated by

, , 3. 45

, , 3. 46

The one of the important aspect of the motion equation can be regarded as friction forces on
the piston (Ilango & Soundararajan, 2012). The common friction model is given by the
Stribeck friction curve (after Stribeck, 1902) and a typical curve can be seen in Figure 3.6. This
curve models the friction forces as a function of the speed and the formulation is given as

3. 47

The characteristics of the curve are determined by three parameters: viscous friction , static
friction , and Coulomb friction . The parameter σ is the viscous friction, is the
parameter for Coulomb friction, and (known as Stribeck velocity) are the parameters for
static friction.

P a g e 28
P a g e 29
 4 Simulation models
This chapter firstly characterizes and interconnects the subsystems of Delaval’s robot arm and
then presents a nonlinear model built upon the relevant mathematical modeling knowledge
given in the previous chapters. Further model simplifications are also proposed and several
linear models are derived. Additionally, determination of the specific model parameters is
represented for a complete understanding of the process.

4.1 Characterization and interconnection of subsystems

The complexity of the modeling of system can be reduced by breaking it into a number of
subsystems (see Figure 4.1).

Figure 4.1. Subsystems of hydraulic servo-systems with interconnections’

I. Depending on the actuator pressures, the valve converts the control signal into oil flow.
The valves are developed to be fast and assumed to show linear input-output behavior;
however, these devices’ actual behavior is usually not ideal, which is also the case with
the valve used in Delaval’s robot arm. This non-ideal behavior of the valve propagates
through the entire robot arm since the valve flow drives the actuator. It must be pointed
out that this is the main reason why the valve, in the modeling process, is explicitly
considered as a separate system.
II. The hydraulic actuator draws oil flows QA and QB and provides mechanical
displacement to the robot arm to carry out the assignment presented in section 1.1.1.
The outputs of the actuator are given as actuator pressures pA and pB, and position p or

P a g e 30
 velocity p while the only input is the external force Fext. Fext could be any unexpected
event that may lead additional force on the arm.
The inertia (impedance) of the actuator base is developed large enough to avoid
parasitic energy exchange between the base and the actuator. This effectively means
that no parasitic motion of the base can occur.
III. If there is a long actuator stroke and relatively long pipelines, pipelines between the
valve and the actuator should be considered as a third subsystem. This is simply
because of the compressibility and the inertia of the oil; pressure waves travel with a
velocity inside the pipelines and are almost ideally reflected at the end of the line.
Because of these dynamics, pressures and flows at the cylinder side must be separated
from those at the valve side (can be observed in Figure 4.1 since these parameters
named with different indices at each side). However, it is not the case in VMS system
and this phenomena is neglected.
IV. Similar to the pipelines between the cylinder and valve, the pipelines between the
power supply system and the valve can be additionally considered as a fourth
subsystem in case the distance between main line and accumulator/valve is large.
However, in Delaval’s hydraulic system, the valves are placed relatively closer to the
cylinder.
V. The power supply cannot maintain the constant pressure flow.

It should be also stressed that the pipeline dynamics can be neglected in case of low-frequency
behavior since they do not play a significant role in the input/out behavior. Therefore, the
model of the hydraulic servo-system is simplified into two subsystems (valve + cylinder) with
a non-ideal power supply.

4.2 Structured nonlinear model

In this section, a reasonably simple model that includes most of the dynamic and non-linear
effects that are relevant to a hydraulic servo system will be presented.

4.2.1 Assumptions
The assumptions are made to simplify the derivation of the non-linear model. These
assumptions are

 Turbulent flow occurs through the flow of the valve.

 The inefficient volumes such as pipelines between the valve and the actuator are
included in relevant cylinder chambers as inefficient volumes.
 The friction model includes coulomb, static and viscous friction (e.g. Stribeck model)
 The surroundings of the actuator and the load are rigid.

P a g e 31
  Any possible dynamic behavior of pressure in the pipelines between valve and actuator
is assumed to be negligible (i.e., valve is located on the actuator).

4.2.2 The model description

A clear overview of the proposed non-linear model is shown in Figure 4.2. It can be seen that
the model consists of three different modules given as valve dynamics, pressure dynamics and
the mechanical part. It should be noted that FL is the load force, Ff is the friction force and Fext
is any external force acting on the cylinder.
For positive control input signal 0, the pressure dynamics model takes the form of

,
4. 1
∝ ,

and for negative control input signal 0

,
4. 2
∝ ,

These equations can be further simplified by assuming constant bulk modulus and fixed
chamber volumes; however, this leads to additional errors. Therefore, further simplifications
are avoided and the pressure dynamics model will be kept as it is given.

Figure 4.2. Structured model of the proposed hydraulic servo-system

P a g e 32
Firstly, the valve dynamics module can be examined. The valve dynamics uses a first order
model to convert the reference signal to actual valve opening. This is given by

1
4. 3

where is given as the reference input of the valve.

However, it should be remembered from section 3.2.1 that the valve coefficient is dependent of
input signal, flow direction and the pressures xv(u*,P, xflow); hence, these non-linear effects are
also included in the valve module.
Finally the mechanical module is formulated

1 4. 4
∝

It should be stressed that this model maintains a reasonably well described non-linear hydraulic
servo model with practical assumptions (see section 4.3.1).
It is also possible to further develop linearized and even more simplified models from the given
non-linear model. These will be investigated further in this chapter.

4.3 Structured linearized model

The linearization of the model starts with linearization of the pressure–flow equations of the
valves 2 . Reformulation of (3.18) with help of Taylor series at a particular operating point
P0=(xv0, pA0, pB0)3

∆ | ∆ | ∆ ⋯ 4. 5

results in

∆ , ∆ , ∆ 4. 6

∆ , ∆ , ∆ 4. 7

By using (4.6) and (4.7); the flow gains can be defined as

2
More detailed information on approximation of non-linear systems can be found in Appendix A.
3
The deviation from the operation point at which the model is linearized is symbolized by the operator “Δ”.

P a g e 33
 ,
, | 4. 8
,

,
, | 4. 9
,

In the same way, the flow-pressure coefficient can be defined as

,
2
, | 4. 10
,
2

,
2
, | 4. 11
,
2

Finally, the pressure sensitivity coefficient can be given as

4. 12

The coefficients , and are regarded as the valve sensitivity coefficients and are
specifically important in determining stability, frequency response, and other dynamic
characteristics (Merritt, 1967).
In particular, the flow gain influences the open loop gain constant in the system; hence, plays
an important role on the system stability. The flow-pressure coefficients influence the damping
ratio of the valve/cylinder system. The valves often have quite large pressure sensitivity
coefficient which presents ability of valves to breakaway large friction load with a relatively
small error (Jelali & Kroll, 2004).

4.3.1 Assumptions
The linearized model has the same assumptions as given in section 4.2.1. Besides, the
linearized model neglects the influence of the change of piston position ∆ on volumes of the
chambers A and B (e.g., and are constant), as well as change of pressures in
effective bulk modulus (i.e., and are constant).

P a g e 34
4.3.2 Model description
Most parts of the non-linear model are kept except some necessary simplifications of non-
linear effects.
For positive control input signal 0, the pressure dynamics model takes the form of

∆ ∆ ∆
∆
∆
∝ ∆ ∆ ∆

, ∆ , ∆
4. 13
, ∆ , ∆

The valve dynamics module uses a first order model to convert the reference signal to actual
valve opening and ignores all other nonlinearities.
Finally the mechanical module can be formulated as

∆
∆
1 4. 14
∆
∆ ∝∆ ∆ ∆

It can be seen from the descriptions that the linear model is very similar to the given non-linear
model except for the influence of nonlinearities. The influence of these effects will be further
investigated in Chapter 5.2.

4.3.3 Parameter feedback

The regular linear model ignores the nonlinearities in the model by applying constant
parameters. This can be applicable for most of the parameters; however, the abnormalities
cannot be ignored.
In this context, it is possible to further improve the linear model by implementing a parameter
feedback. It should be noted that the given models up to now are solved with discrete solvers.
This brings the opportunity that each constant parameter actually can be updated based on a
feedback of the latest state of the system. This approach can be better understood from Figure
4.3.

P a g e 35
Figure 4.3. a) The solver algorithm of the linearized model b) The solver algorithm of the linearized model with a
parameter feedback

It is clearly seen that the linearized model neglects the influence of the change of piston
position ∆ on the volumes of the chambers A and B as well as the change of pressures in
effective bulk modulus. However, the linear model with a parameter feedback updates these
parameter every time step based on the state of the system. Therefore, a relatively more
accurate model with further considerations on non-linearities can be implemented by using this
approach.

4.4 Simplified linearized model

The purpose of the simplified model is to further decrease the dependencies of the pressures on
each chamber. In this sense, a linear relationship between the chamber pressures and the load
pressure are given as4

1
∆ ∆ 4. 15
1

∝
∆ ∆ 4. 16
1

At this point, if the load pressure is rewritten and differentiated;

∝ 4. 17

∝ 4. 18

Now, combining equation (4.15-4.18) and (3.38-3.39) the pressure dynamics equation becomes

4
For further information on linearization process, see Appendix B.

P a g e 36
 1
∆ ∆ ∆ 4. 19

The parameters , , in (4.19) can be expressed as

, , 4. 20

4. 21

and

1
4. 22
, 1 , 1
∝
1 1

Finally equations (4.17-4.18) and (3.43) are combined to give the motion equation (mechanical
module)

∆
∆
1 4. 23
∆
∆ ∆ ∆

It should be stressed that this approach reduces the state of the system by one with the help of
developing a relationship between each chamber. The model neglects the influence of the
change of piston position ∆ on volumes of the chambers A and B as well as change of
pressures in effective bulk modulus.

4.4.1 Parameter feedback

It is also possible to further improve the simplified linear model by parameter feedback
implementation as it is done in the regular linear model. Similarly, each state parameter will be
updated based on a feedback that delivers the latest state of the system. For further information,
see section 4.3.3.

4.5 Further information on essential model parameters

Some of the crucial parameters can be estimated by using manufacturer’s catalogue
information and drawings. If the parameters cannot be found on those sheets, they can be

P a g e 37
calculated either by using data from an experiment or by an expression provided in the
literature.

4.5.1 The valve flow-signal function

If the supply pressure is constant, it can be seen that a valve is partially characterized by the
relationship between valve input u* and valve flow Q, when the flow is in steady state. This
relationship can be used to determine flow-signal function Q(u*). Proportional valves are
characterized by straight proportional lines in ideal cases. In practice, these curves deviate from
ideal cases due to saturation for larger valve inputs, valve overlap/underlap, valve hysteresis
and some non-smooth behavior around zero input. A typical flow-signal curve can be seen in
Figure 4.4. It should also be noted that this curve is used for the calculation of the reference
flow coefficient.

Figure 4.4. The flow-signal curve for the valve implemented in the developed model (Parker, 2013)

4.5.2 Valve overlap and underlap

This parameter is generally provided clearly in the manufacturer’s catalogue information.
Moreover, the valve overlap can easily be determined by using Figure 4.4, i.e., the beginning
of the fluid flow for the smallest valve input signal.

4.5.3 Actuator dimensions and mass

The manufacturer’s construction drawings of the actuator often precisely provide required
piston areas and stroke thus, the actuator dimensions and the mass can easily be determined by
using these sheets.

P a g e 38
It is relatively more difficult to determine the inefficient volumes of the actuator such as
pipelines between valves and actuator, use of safety manifolds etc. These parameters can be
determined by relatively less accurately calculations with the help of using construction
drawings.
The most straight forward provision of the masses of the components can be done by
weighting the objects. However, less precise estimations can also be made by using
construction drawings.

4.5.4 Friction forces

As it is introduced in Section 3.2.3.2, non-linear friction forces in an actuator is one of the main
source of the nonlinearities in the cylinder model.
As proposed in the given section, it is generally effective enough to use the Stribeck friction
model. In this model, viscous friction is dependent on the velocity of the piston and plays an
effective role on the stability and damping of hydraulic system. Thus, viscous friction should
never be omitted in the model.
Viersma (1980) argues that Coulomb friction is the magical factor that is responsible for the
most of the non-linearity in the actuator and is always suspected when something unexpected
happens. Jelali & Kroll (2004) also conclude that Coulomb friction may cause instability issues
in practice, i.e., stick-slip phenomena at low speeds.
In this context, the friction forces can be experimentally calculated by measuring the pressures
of each cylinder chamber and using

∝ 4. 24

where pA and pB are pressures in chambers A and B respectively. It should be noticed that the
load and any external forces are neglected in this formula. Moreover, if the data would be
taken in a steady state condition, the acceleration term disappears and (4.24) becomes:

∝ 4. 25

It also possible to obtain the data for the friction forces from the manufacturer’s data sheets
whereas some of the parameters of the Stribeck friction model can be found in the literature
and be adapted to the model used.

4.5.5 Leakage coefficients

The external leakage flow from actuator to casing is neglected; therefore, only the internal
leakage flow needs to be determined.
The internal leakage coefficient CLi can be obtained either by manufacturer’s data sheets or by
an experimental setup. The latter is done by fixing the piston position rigidly and taking

P a g e 39
pressure measurements from each chambers of cylinder. This experimental setup is based on
the fact that a small valve opening would lead to supply pressure on the one side of the
chamber. However, in case of leakage flow, there will be oil flow and this would cause a
pressure drop in the relevant chamber. In this sense, the pressure in the given chamber would
be less than the supply pressure while the pressure in the other chamber would equal to the
tank pressure since the valve enables free flow of small amounts of oil (see flow equation in
chapter 3.2.3.1).

4.6 Simulation package

The simulation package Matlab R2011b has been selected even though literature mainly
proposes Matlab/SIMULINK. In house Delaval experts conclude that Matlab script models are
often more flexible since the user can adapt any part of the script with any other Matlab or
other software in DeLaval. Moreover, the script models often provide more transparency to its
user.

P a g e 40
 5 Results
The chapter starts with a description of the experiment carried out to acquire measurement
data for the validation. The model results are obtained with the methods described in the
previous chapters. They are compiled, analyzed and compared with the experimental data. To
acquire complete understanding of the model dynamics, further sensitivity analysis is also
carried out.

5.1 The introduction of measured data

The validation of the simulation studies often requires experimental data; hence, an
experimental setup is built at DeLaval.
In the experiment, pressure in both chambers, the pressure at the pump outlet, the displacement
of the rod, the valve input signal and the elapsed time are measured. The measured data and
measurement frequency for each variable are shown in Table 5.1.
Table 5.1. The measured data and measurement frequency for each variable

Measurement
Measured quantity Nomenclature
Frequency (Hz)
Chamber Pressure A pA 40
Chamber Pressure B pB 40
Pump Outlet Pressure PS 40
Displacement xp 200
Valve input signal u 200
Elapsed time t 200

It is assumed that the robot arm only moves in the X and Y direction and maintains the motion
always parallel to the ground. In other words, there are no gravity forces acting on the cylinder
in the model. However, it is unavoidable that small deviation in Z axis occurs. Even though,
the gravity is not included in the models, the gravity leads to deviations on the chamber
pressures on the experimental data which will be further discussed in section 6.1. Besides, the

P a g e 41
displacement is estimated with the help of the measured angle between the robot arm and the
actuator rod; thus, deviation in Z axis direction leads to additional errors in the calculation of
the piston position. However, the errors in the displacement can be compensated since the
starting position and final position of the arm is known.

5.1.1 Friction forces

One of the crucial aspects of the model is the simulation of the friction forces. However, the
verification of the proposed model (3.47) should be done by a comparison with experimental
data. For this purpose, the friction forces are calculated by using the experiment proposed in
section 4.5.4 and the experimental friction forces are estimated for different velocities with the
help of (4.25). The comparison of values acquired by the experiment and the friction model is
given in Table 5.2. It can be easily seen that the friction model estimates the related friction
forces with accuracy that has a relative error smaller than ±3.6%.

Table 5.2. Verification of friction model by comparison of the simulated data with the experimental data

Experimental vs simulated friction forces

Experimental Velocity Simulated Relative
Friction Force (N) (m/s) Friction Force (N) Error
120.53 0.165 124.20 -3.04%
113.7 0.123 112.44 1.11%
103.52 0.078 99.84 3.55%
94.75 0.048 91.48 3.45%
85.3 0.022 84.71 0.69%
79.86 0.01 82.64 -3.48%
Friction model parameters
cS Fv Fc0 Fs0
0.01 280 78 5

5.1.2 The flow coefficient

It is discussed in section 3.2.1 that the flow coefficient often deviates from the theoretical value
for different valve openings. This effect is investigated by building an experimental setup and
calculating the flow coefficient by using data obtained from the experiment with the help of
(3.18). In the experiment, the supply pressure (pS), chamber pressures (pA, pB) and position of
the rod (xp) are measured. By using this data, experimental flow coefficient value is calculated.
It should be pointed out that, due to valve hysteresis, this calculation is made only for the
positive displacement (see section 3.2.1.1). The estimated compensation parameter is
shown in Figure 5.1.

P a g e 42
 Figure 5.1. Deviation of compensation parameter for measured data

The valve acts relatively proportional between the valve input signals 0.2 to 0.5; while, its
behavior is dramatically altered in other regions. The lower flow coefficient means that less
flow through the valve would occur than the theoretical estimation.

5.1.3 The valve hysteresis

It is given in section 3.2.1, that the valves often have hysteresis that should not be omitted in
the mathematical models. This phenomenon is investigated by comparing the flow coefficients
for both directions for the same input signal. The estimated valve hysteresis compensation
parameter , can be seen in Figure 5.2.
Figure 5.2 clearly shows that the valve hysteresis is around 5% most of the time. However, the
hysteresis varies between 0-12.5 % for the linear region and shows nonlinear behavior.

P a g e 43
 Figure 5.2. . Deviation of valve hysteresis compensation parameter ,

5.2 Comparison of models

In this section, a comparison of the nonlinear, linear and simplified linear models is carried out.
The measured data for the given valve input signal is also included. Furthermore, it should be
stressed that the compensation parameters ( and , ) are omitted in all models for
a fair comparison, although, it can be easily implemented in nonlinear model and with the help
of a feedback loop for linear models.
The comparison is made for various step input signals because of the nonlinear behavior of
hydraulic system and can be seen in Figure 5.3 to Figure 5.7.
It can be seen from Figure 5.3 that none of the models can predict the displacement accurately
for small openings. However, the nonlinear model shows better pressure prediction than linear
model. Between the valve opening region of 25.8% to 50.5% (see Figure 5.5 and Figure 5.6),
the linear model shows the highest accuracy on the displacement and pressures. For the large
valve opening (see Figure 5.7), the models completely diverge from the experimental data for
the displacement and the chamber pressures.

P a g e 44
Figure 5.3. The comparison of models for the input signal of 20.3%

P a g e 45
Figure 5.4. The comparison of models for the input signal of 25.8%

P a g e 46
Figure 5.5. The comparison of models for the input signal of 36.8%

P a g e 47
Figure 5.6. The comparison of models for the input signal of 50.5%

P a g e 48
Figure 5.7. The comparison of models for the input signal of 91.6%

P a g e 49
5.3 Comparison of improved models
In this section, a comparison of the improved nonlinear, linear and simplified linear models is
carried out. In other words, the improved models include compensation parameters ( and
, ) for valve hysteresis and valve flow nonlinearities. Moreover, the parameter
feedback also updates the state variables at each time step (see section 4.3.3).
The measured data for the given valve input signal is also included. The comparison is made
for various step input signals because of the nonlinear behavior of the hydraulic system and
shown in Figure 5.8 to Figure 5.13.
It can be seen from Figure 5.38 that the nonlinear model can predict the displacement with a
better accuracy for small openings. However, it should be noted that the accuracy of improved
models are better than regular models. Between the valve opening region of 25.8% to 50.5%
(see Figure 5.9 and Figure 5.12), the nonlinear model show the highest accuracy on the
displacement and pressures. Similarly, the accuracy of improved models is better than regular
models in this region, too. For the large valve opening (see Figure 5.13) nonlinear model is still
able to forecast displacement and pressures, while other models are not. Although, they are
more accurate than regular models.

5.4 Comparison of all proposed models

In this chapter, the steady state displacement errors of all models are compared in Table 5.3 to
provide a complete overview to the reader. It should be given that relative errors below than
7.5% is regarded as highly accurate, between 7.5 % an 20 % is given as moderately accurate
while relative errors larger than 20 % are assumed to be inaccurate.

Table 5.3. Relative displacement errors of each model

Valve Nonlinear Improved Linear Improved Simple Simple

Signal Nonlinear Linear Linear Imp. Linear
20.3% Inaccurate Accurate Inaccurate Inaccurate Inaccurate Inaccurate
25.8% Inaccurate Accurate Moderate Inaccurate Inaccurate Moderate
36.8% Inaccurate Accurate Accurate Accurate Inaccurate Accurate
50.5% Inaccurate Accurate Accurate Accurate Inaccurate Moderate
71.3% N/A Accurate N/A Accurate N/A Moderate
91.6% Inaccurate Accurate Inaccurate Moderate Inaccurate Moderate

P a g e 50
Figure 5.8. The comparison of models for the input signal of 20.3%

P a g e 51
Figure 5.9. The comparison of models for the input signal of 25.8%

P a g e 52
Figure 5.10. The comparison of models for the input signal of 36.8%

P a g e 53
Figure 5.11. The comparison of models for the input signal of 50.5%

P a g e 54
Figure 5.12. The comparison of models for the input signal of 71.3%

P a g e 55
Figure 5.13. The comparison of models for the input signal of 91.6%

P a g e 56
5.5 Sensitivity Analysis
In this section sensitivity analysis is carried out to investigate how the uncertainty in the output
of the model can be apportioned to different sources of uncertainty in its inputs. It should be
given that nonlinear model is used for sensitivity analyses. And while the other parameters of
the simulation could be determined easily, there were uncertainties on the flow coefficient, the
valve response time, supply pressures, the mass variation and external forces, therefore they
will be investigated further.
The main purposes of the analysis are:

 Testing the robustness of the results since the presence of uncertainty in given
parameters is unavoidable in the model.
 Reducing the uncertainty by detecting which of these parameters cause substantial
influence in the output and should therefore be the focus of attention if the robustness
were to be improved
 Proposing further model simplification by avoiding parameters that have no effect on
the output (e.g. assuming constant supply pressure).

5.5.1 Influence of flow coefficient

Flow coefficient (cv) is one of the most crucial parameter in the simulations and derived from
the valve catalogue provided by suppliers and modified by using experimental data if it is
needed in the simulations. It mainly acts as an open loop gain in the flow models and has a
crucial role on the stability of the system. Thus, it becomes the most crucial variable in the
flow equation even though it is sometimes extremely hard to estimate it precisely. Hence, it is
very obvious that carrying out a sensitivity analysis on the flow coefficient and revealing
possible outcomes are essential for the sake of the simulations. The sensitivity analysis for the
flow coefficient is shown in Figure 5.14.

5.5.2 Influence of valve response time

It is fair to say that valve response time is relatively less crucial but it is still interesting to see
if it affects the simulation in any unpredicted way. It must be stated that valve response time
can be effected by many unknown conditions and sometimes becomes more stochastic than
other deterministic parameters (see section 3.2.1) and carried out sensitivity analysis is
presented in Figure 5.15.

P a g e 57
5.5.3 Influence of supply pressures
In house DeLaval experts presented their opinion on the supply levels and the opinions
suggested that the pressures can vary between 25-35 bars in the VMS. Therefore, it is essential
to include this parameter in the sensitivity analyses, as shown in Figure 5.16, to increase
reliability and practicality.

5.5.4 Influence of mass variation of the system

Another suggestion by in house experts of Delaval proposes that mass can vary based on
various external sources and/or design variations. Therefore, it is also believed that a sensitivity
analysis should be carried out on this parameter and can be seen in Figure 5.17.

5.5.5 Influence of possible external forces

The main purpose of this is to reveal the feasibility of simulation in the occurrence of any
external forces (Fext) and can be observed in Figure 5.18.

5.6 Further concerns on effective empirical bulk modulus

There are various effective empirical bulk modulus models proposed in the literature.
Moreover, each proposal points out its own advantages; hence, it is interesting to examine the
effect of different bulk modulus models for the sake of the simulation. Moreover, the
theoretical cases where the bulk modulus is relatively large or small will be also investigated
here as can be seen in Figure 5.19.

P a g e 58
Figure 5.14. Influence of the flow coefficient on the outputs

P a g e 59
Figure 5.15. Influence of the valve response time on the outputs

P a g e 60
Figure 5.16. Influence of the supply pressure on the outputs

P a g e 61
Figure 5.17. Influence of the mass variation on the outputs

P a g e 62
Figure 5.18. Influence of the parasitic forces on the outputs

P a g e 63
Figure 5.19. The influence of the bulk modulus

P a g e 64
 6 Discussion and Conclusions
In this chapter, the obtained results are further investigated and the crucial points are
discussed. Based on the discussion, the most critical conclusions are stressed. Any possible
future work is also explained in this chapter.

6.1 Discussion
Firstly, the implemented friction model predicts the friction forces with relatively small errors.
This may occur due to the piston rod moves with relatively small velocities and the cylinder
has been produced with small tolerances.
Even though, proportional valves often act proportionally to the input signal, it can be seen that
that is not entirely correct with the current proportional valve. However, it must be stated that
the VMS robot arm operates under 0.5u* region5 and these nonlinearities are not encountered
in the application. Moreover, valve hysteresis is another aspect of the model as it shows also
nonlinear behavior and an additional attention must be paid to the determination of this
parameter to accurately predict the displacement for both directions.
For the small valve input signal (u*=20.3%), none of the models accurately predicts the
displacement although the linear and the nonlinear models provide more accurate simulations
than the simple linear model. It can be seen that the nonlinear model can more accurately
predict the chamber pressures while the linear model fails to do so (see Table 5.3). One remark
should be made on the chamber pressures on the experimental data. The gravity forces affect
the arm once the arm deviates from the starting position. Since the gravity forces are changing
based on the displacement, this can clearly be observed on the pressures. In other words, there
is an increasing force applied on the piston due to the gravity. The trend occurs simply because
larger angle means that larger gravity force on the arm. When the arm stops, a steady state
pressure difference between chamber A and B occurs to counteract the gravity. It is also the
reason of pressure difference between the nonlinear model and the experimental data since the
nonlinear model does not include the influence of the gravity.

5
It must be noticed that the input valve signal is normalized to 1.

P a g e 65
The valve behavior can be assumed to be linear in case valve input signal varies between 26%
and 51%. Generally, the accuracy of the models is increased in this area, especially with the
linear model since there are less nonlinear effects. The nonlinear model generally predicts the
pressure levels accurately while the linear model predicts both values accurately enough.
However, for models in large openings region (u*=91.6%), the displacement cannot be
predicted by none of the models even though the chamber pressures are accurate enough.
The improved nonlinear model can predict both displacement and the chamber pressures
accurately for all regions. Moreover, all of the improved models can predict both displacement
and the chamber pressures more accurately except for the small openings. Especially, the
simple linear model also predicts the displacement relatively accurate enough while it fails to
do so when a regular simple linear model is used.
It is clear that the flow coefficient has a significant impact on the displacement of the arm,
while it has no impact on the steady state pressure chambers. On the other hand, the supply
pressure has a remarkable influence on both steady state chamber pressures and the
displacement. The valve response time alters the time the system reaches the steady state. Both
the mass variation and external forces do not have major impact on the displacement.
However, the mass variation affects the dynamics of the system significantly. Larger mass
leads to higher pressure level on the chamber A which is needed to overcome inertial forces.
Moreover, the external forces affect both the dynamic and steady state chamber pressures.
While two different empirical models suggest similar behavior, increased bulk modulus values
shows that the system would have reached steady state faster or vice versa.

6.2 Conclusions
It is clear from the discussion and Table 5.3 that none of the regular models are accurate
enough to simulate the robot arm for various valve input signals. However, the linear model is
accurate enough to simulate both chamber pressures and the displacement for the linear region
and can be used to simulate only this region. Besides, since the nonlinear model is able to
predict pressure levels more accurately for all regions; it must be used for prediction of
pressures if it is necessary.
It is fair to say that the improved nonlinear model is safe to use for all valve openings and
inferior to any other models. Clearly, the parameter feedback makes both the linear and the
simplified linear models more reliable. However, it should be understood that the linear model
contains more equations and require more calculating power than the nonlinear model if the
parameter feedback is implemented. Therefore, even though the linear model is accurate
enough, using the nonlinear model is still more logical for practical purposes (e.g. need for less
calculation power). On the other hand, the simplified linear model requires less calculation
power and can be considered for specific applications. However, the simplified linear model

P a g e 66
does not provide the chamber pressures and this can be a drawback if those variables need to
be estimated.
Sensitivity analysis points out that the flow coefficient is the most crucial parameter for the
estimation of the displacement of the valve and must be thoroughly investigated. Even though,
most of the literature claims that implementing a constant supply pressure is a valid
assumption, sensitivity analysis on this parameter stresses that such an assumption may cause
major errors in both chamber pressures and estimated displacement in the simulations. In this
context, it must be given that the measured supply pressures are varying in a significant range
in the DeLaval VMS system.
The dynamics of the system is affected by the mass variation and the external forces; though,
these parameters must be estimated carefully if dynamics of the system has the highest priority.
Moreover, the valve response time can be crucial for fast dynamics and large valve response
time value must be avoided for such cases. Furthermore, the bulk modulus is another important
parameter that alters the system dynamics and the most appropriate bulk modulus model must
be carefully selected by investigating the system properties. Then the closest empirical model
can be picked with help of the literature.

6.3 Future work

It is stated that the models do not implement the effect of the gravity; although it was
unavoidable in the experiment. Therefore, a gravity model can be further introduced for
increased accuracy.
For the inaccurate models, the system state can be updated by using additional considerations
(e.g. Kalman Filter)
More smooth description of the flow coefficient can be made by operating the robot arm for
more points. This may also increase the robustness of the models.
The flow though the valves can be measured with flow sensors and this may help cross-
checking some of the parameters used in the simulation. It should be noted that the flow
coefficient is calculated indirectly since there is no measurement of flow through the valve
available.
The DeLaval VMS robot arm contains three hydraulic actuator that drives two links. Therefore,
these actuators can be connected with a more detailed mathematical model to provide complete
description of the robot arm.

P a g e 67
 7 References
Bar–Meir, G. (2011). Basics of Fluid Mechanics. Boston, MA 02110-1301 USA: Free
Software Foundation, Inc.
Cundiff, J. S. (2002). Fluid Power Circuits and Controls: Fundamentals and Applications.
Boca Raton, Florida 33431: CRC Press LLC.
Dasgupta, K., & Murrenhoff, H. (2011). Modelling and dynamics of a servo-valve controlled
hydraulic motor by bondgraph. Mechanism and Machine Theory, 46(7), 1016-1035.
Dasgupta, K., & Watton, J. (2005). Dynamic analysis of proportional solenoid controlled
piloted relief valve by bondgraph. Simulation Modelling Practice and Theory, 13(1),
21-38.
DeLaval. (2013, 10 15). About DeLaval. Retrieved from Delaval:
http://www.delaval.com/en/About-DeLaval/
Eggerth, S. (1980). Beitrag zur messung von volumenströmen viskoser flussigkeiten in
druckleitungen. Diss: Technical University of Dresden.
Eryilmaz, B., & Wilson, B. H. (2006). Unified modeling and analysis of a proportional valve.
Journal of the Franklin Institute, 343(1), 48-68.
Fisher. (2013, 10 22). Control valve handbook. Retrieved from Columbia National University:
http://www.docentes.unal.edu.co/sorregoo/docs/Control%20Valve%20Handbook.pdf
George, H., & Barber, A. (2007). What Is Bulk Modulus, and Why Is It Important? Hydraulics
& Pneumatics.
Goodson, R. E., & Leonard, R. G. (1972). A Survey of Modeling Techniques for Fluid Line
Transients. J. Basic Eng, 94(2), 474-482.
Gordic, D., Babic, M., & Jovicic, N. (2004). Modelling of spool position feedback servo-
valves. International Journal of Fluid power, 1(5), 37–50.
Ilango, S., & Soundararajan, V. (2012). Introduction to Hydraulics and Pneumatics (Vol.
Second Edition). New Delhi: PHI Learning Private Limited.

P a g e 68
Jelali, M., & Kroll, A. (2004). Hydraulic Servo-systems: Modeling, Identification and Control.
London: Springer-Verlag London Limited.
Johnson, J. L. (2013, 10 25). How to interpret valve specifications. Retrieved from Hydraulics
and pneumatics: www.hydraulicspneumatics.com/content/site200/articles
Katz, J. (2010). Introductory fluid mechanics. New York: Cambridge University Press.
Kilic, E., Dolen, M., Koku, A. B., Caliskan, H., & Balkan, T. (2012). Accurate pressure
prediction of a servo-valve controlled hydraulic system. Mechatronics, 22(7), 997-
1014.
Kundu, P. K., & Cohen, I. M. (2008). Fluid Mechanics (Fourth Edition ed.). Kidlington,
Oxford: Elsevier.
Kundu, P. K., Cohen, I. M., & Dowling, D. R. (2011). Fluid Mechanics (Vol. 5). Amsterdam:
Elsevier.
Lee, K.-I., & Lee, D.-K. (1990). Tracking control of a single-rod hydraulic cylinder using
sliding mode. 29th SICE Annual conference, (pp. 865-868). Tokyo, Japan.
Maneetham, D., & Afzulpurkar, N. (2010). Modeling , simulation and control of high speed
nonlinear hydraulic servo system. World Journal of Modelling and Simulation, 6(1),
27-39.
Maria, A. (1997). Introduction to modeling and simulation. Winter simulation conference, (pp.
7-13). S. Andradottir.
Menshawy, M. T. (2006). Investigation of dynamic behavior of an electrohydraulic
proportional system. Cairo, Egypt: Military Technical College.
Menshawy, T. M., Moghazy, M. A., & Lotfy, A. H. (2009). Investigation of Dynamic
Performance of an Electro-Hydraulic Proportional System. 13th International
Conference on Aerospace Sciences & Aviation Technology (pp. 1-18). Cairo, Egypt:
ASAT.
Merritt, H. (1967). Hydraulic control systems. New York: John Wiley & Sons, Inc.
Mitianiec, W., & Bac, J. (2011). Mathematical model of the hydraulic valve timing system.
Journal of KONES Selected full texts, 18(2), 311-321.
Norvelle, F. D. (1999). Electrohydraulic Control Systems. Richmond, TX, U.S.A: Prentice
Hall.
Novak, P., Guinot, V., Jeffrey, A., & Reeve, D. E. (2010). Hydraulic Modelling – Principles,
methods and applications. New York : Spon Press.

P a g e 69
Parker. (2013). Catalog HY14-2550/US. Technical Information: Direct Operated Proportional
DC Valves (Series D1FB and D3FB). Elyria, , Ohio, USA: Parker Hannifin
Corporation.
Tenali, V. S. (2008). Simualatıon of Electro-hydraulic Servo Actuator. Orissa: Department of
Mechanical Engineering National Institute of Technology, Rourkela.
Totten, G. (2000). Handbook of hydraulic fluid technology. New York, NY: Marcel Dekker,
Inc.
Viersma, T. (1980). Analysis, Synthesis and Design of Hydraulic Servosytems and Pipelines.
Elsevier Science Ltd.
Walters, R. B., Eng, C., & Mech, F. I. (1991). Hydraulic and Electro-hydraulic Control
Systems. New York: Elsevier Applied Science.
Wang, J., Gong, G., & Yang, H. (2008). Control of Bulk Modulus of Oil in Hydraulic Systems.
International Conference on Advanced Intelligent Mechatronics (pp. 1390-1395).
Xi'an, China: IEEE/ASME.
Wei, J., & Xianxiang, H. (2010). The modeling and parameter identification of the hydraulic
proportional valve. 201O International Conference on Computer, Mechatronics,
Control and Electronic Engineering (CMCE), (pp. 482-485).
Wijnheijmer, F. P. (2005). Modelling and control of a hydraulic servo system H ∞ control and
LPV control versus classical control. Eindhoven: University of Technology Eindhoven.
Younkin, G. W. (2003). Industrıal servo systems: Fundamentals and applications. New York,
NY : Marcel Dekker, Inc.

P a g e 70