LEARNING UNIT 2

THEORETICAL MODELS IN CHEMICAL ENGINEERING

CONTENTS PAGE

LEARNING OUTCOMES ............................................................................................ 11

2.1. INTRODUCTION ............................................................................................... 13
2.2. MODELLING PRINCIPLES ................................................................................. 14
2.2.1 Introduction............................................................................................................................ 14
2.2.2 Degrees of freedom ............................................................................................................... 15
2.2.3 Mass and energy balances – recap ........................................................................................ 16
2.3. MODELLING OF A SIMPLE EXAMPLE .............................................................. 17
2.4. LINEARISATION OF NON-LINEAR SYSTEMS .................................................. 22
2.4.1 Linearization using Taylor series ............................................................................................ 22
2.4.2 Deviation variables ................................................................................................................. 25
2.4.3 Linearisation of systems with many variables ....................................................................... 27
2.5. CONCLUSION .................................................................................................... 28
2.6. PROBLEMS ........................................................................................................ 29
2.7. SOLUTIONS TO PROBLEMS............................................................................. 32

10
PRC3701 LEARNING UNIT 2
 LEARNING OUTCOMES
After completion of this learning unit, you should be able to:
• Describe the procedure for developing a mathematical model of a dynamic
process
• Derive theoretical models for selected and simple chemical engineering
processes
• Determine the behaviour of the system on the basis of the model, and interpret
the results
• Describe how you can use the concept of degrees of freedom and how you can
calculate the degrees of freedom
• Identify non-linear equations or functions
• Describe how you can linearise non-linear systems
• Apply linearisation principles to selected and simple chemical engineering
processes to linearise equations or functions containing non-linear equations

This learning unit is based on the following chapter in the prescribed textbook:

o Chapter 2: Theoretical Models of Chemical Processes

Consult the textbook if you need extra help in understanding some of the concepts in
the study guide. Also, follow the links to the eight video clips below:

• Mathematical Modeling – Material Balances:

https://www.youtube.com/watch?v=2eGzc1ty-qQ
• Mathematical Modeling – Energy Balances:
https://www.youtube.com/watch?v=WThXhZ4mBz8
• Mathematical Modeling – Multiple Balances:
https://www.youtube.com/watch?v=9eQWBmURUik
• Introduction to Degrees of Freedom:
https://www.youtube.com/watch?v=tW1ft4y5fQY
• Degree of Freedom – Analysis on a Single Unit:
https://www.youtube.com/watch?v=6Rx2ry1P6ME
• Linearization of Differential Equations:
https://www.youtube.com/watch?v=EJxGZKWnuxQ

11
PRC3701 LEARNING UNIT 2
 • Linearization of Differential Equations:
https://www.youtube.com/watch?v=kd0GlcOpXUA
• First-Order Transfer Function with Linearization:
https://www.youtube.com/watch?v=aQ3N7Dw0L-w

These links are also available on the module site under Additional Resources.

12
PRC3701 LEARNING UNIT 2
 2.1. INTRODUCTION
A theoretical model is a description or representation used to understand the way in
which a particular system or process works. In this learning unit, we will look at how we
devise or create theoretical models that can then be used to control a process.

A model has limitations, based largely on the assumptions we have used to create it,
and so it will never represent the complete reality of the process. For example, if we
assume steady state in creating a model, then the model can only be used for steady-
state operations. If the process is unsteady state, as at the startup or shutdown of a
plant, we will need a different model for that – so this is a limitation. However, even from
an incomplete picture of reality, we may be able to learn several things from the model.

The model we devise or come up with can be tested under extreme

conditions/circumstances. This is an advantage, as it is sometimes difficult to test the
true process or system at these extreme conditions. Models are often "embedded" in
the controller itself to assist in control, in other words, the controller can use a process
model to anticipate the effect of a control action. From our example of taking a shower
in learning unit 1, we need to know whether an increase or decrease in hot water will
affect the warm water we want. An increase in the flowrate of hot water will result in the
warm water becoming hot (positive gain), while a decrease in the flowrate of hot water
will lead to the warm water becoming cold (negative gain). So, if we are to design a
controller, we need to know whether an increase in the manipulated input (hot water)
will result in an increase or decrease of our process output variable (warm water
becoming hot or cold) – that is, we need to know whether the process gain is positive or
negative.

13
PRC3701 LEARNING UNIT 2
 2.2. MODELLING PRINCIPLES
2.2.1 Introduction
There are four steps to follow when devising a control scheme. These are:
Step 1 – Assess the process and define the control objective(s). In carrying out this step
we need to ask ourselves questions like:
1. Why should we consider process control, in other words, what is the need for
process control?
2. Is process control the only answer in this context, or can we solve the problem by
alternative means, such as redesigning the process or part of the process?
3. What is the primary objective of the process control?

Step 2 – Identify the process (input and output) variables that will be used to achieve the
control objective. In carrying out this step we need to ask ourselves questions like:
1. Which output variable(s) that is/are important to achieve our process objective
should we monitor and measure?
2. Which input variable(s) (either manipulated or disturbance) do we need to monitor,
measure and ultimately change (manipulate) to achieve our control objective?

Step 3 – Decide on the control structure to use. In carrying out this step we need to ask
ourselves this question:
What configuration is best suited to the particular problem we are experiencing?
The choice is usually between feedback and feedforward.

Step 4 – Design the controller. Here we need to obtain a model that relates inputs to
outputs using historical information and see whether it is able to predict what may happen
when there are changes. When we are developing a model we need to do two things:
1. We need to clearly state our assumptions, if any. These will be the limitations of
our model when it is in use.
2. We need to consider the model requirements, as these are a function of the end-
use of the model, in other words, what we intend to use the model for. Ultimately
we are interested in designing a controller (feedback or feedforward) and also in
simulating the behaviour of the control system. However, at the moment we do not
14
PRC3701 LEARNING UNIT 2
 have enough knowledge of the mathematics involved, so our objective is to
develop a model that relates the inputs (whether manipulated or disturbance) to
measured outputs that we wish to control.

2.2.2 Degrees of freedom

The concept of degrees of freedom analysis is used in different aspects of chemical
engineering as well as in the development of process control strategies. Writing all the
equations for a process and counting the variables is often a tedious process which is
prone to a lot of errors, so it is important to have a simple but also generic method to
determine the degrees of freedom for control of a whole process. In simple terms,
degrees of freedom (DOF) is defined as:

DOF = Number of variables of the system – Number of equations of the system

In other words, degrees of freedom are the variables that have to be set in order to
have a completely determined system so that when we solve the equations we get
unique solutions.
✓ If DOF < 0, there are more equations than variables, meaning that each or all the
variables will have more than one solution.
✓ If DOF = 0, the number of equation is exactly the same as the number of variables,
meaning that all variables will have unique solutions.
✓ If DOF > 0, there are fewer equations than variables, meaning that we cannot solve
to get unique values for the variables. In other words, we need to make certain
assumptions to some variables.

It is important to determine degrees of freedom of our system so as to see whether we

need to make additional assumptions or increase the number of equations. The model
we need to devise should have unique solutions so that it is able to model our system.
After we have our mathematical model, we want to solve to find the relationship
between the output and the input variables for process control.

15
PRC3701 LEARNING UNIT 2
 2.2.3 Mass and energy balances – recap
From your diploma modules, you know that the mass and energy balance equation is
given as:
𝑅𝑎𝑡𝑒 𝑜𝑓 𝑚𝑎𝑠𝑠/𝑒𝑛𝑒𝑟𝑔𝑦 𝑎𝑐𝑐𝑢𝑚𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑖𝑛 𝑎 𝑠𝑦𝑠𝑡𝑒𝑚 =

𝑅𝑎𝑡𝑒 𝑜𝑓 𝑚𝑎𝑠𝑠/𝑒𝑛𝑒𝑟𝑔𝑦 𝑒𝑛𝑡𝑒𝑟𝑖𝑛𝑔 𝑎 𝑠𝑦𝑠𝑡𝑒𝑚 −

𝑅𝑎𝑡𝑒 𝑜𝑓 𝑚𝑎𝑠𝑠/𝑒𝑛𝑒𝑟𝑔𝑦 𝑙𝑒𝑎𝑣𝑖𝑛𝑔 𝑎 𝑠𝑦𝑠𝑡𝑒𝑚 +

𝑅𝑎𝑡𝑒 𝑜𝑓 𝑚𝑎𝑠𝑠/𝑒𝑛𝑒𝑟𝑔𝑦 𝑔𝑒𝑛𝑒𝑟𝑎𝑡𝑒𝑑 𝑖𝑛 𝑎 𝑠𝑦𝑠𝑡𝑒𝑚 −

𝑅𝑎𝑡𝑒 𝑜𝑓 𝑚𝑎𝑠𝑠/𝑒𝑛𝑒𝑟𝑔𝑦 𝑐𝑜𝑛𝑣𝑒𝑟𝑡𝑒𝑑 𝑖𝑛 𝑎 𝑠𝑦𝑠𝑡𝑒𝑚

When there are no chemical reactions involved, the equation becomes:

𝑅𝑎𝑡𝑒 𝑜𝑓 𝑚𝑎𝑠𝑠/𝑒𝑛𝑒𝑟𝑔𝑦 𝑎𝑐𝑐𝑢𝑚𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑖𝑛 𝑎 𝑠𝑦𝑠𝑡𝑒𝑚 =

𝑅𝑎𝑡𝑒 𝑜𝑓 𝑚𝑎𝑠𝑠/𝑒𝑛𝑒𝑟𝑔𝑦 𝑒𝑛𝑡𝑒𝑟𝑖𝑛𝑔 𝑎 𝑠𝑦𝑠𝑡𝑒𝑚 − 𝑅𝑎𝑡𝑒 𝑜𝑓 𝑚𝑎𝑠𝑠/𝑒𝑛𝑒𝑟𝑔𝑦 𝑙𝑒𝑎𝑣𝑖𝑛𝑔 𝑎 𝑠𝑦𝑠𝑡𝑒𝑚

For steady-state operation, the accumulation term becomes zero. So:

𝑅𝑎𝑡𝑒 𝑜𝑓 𝑚𝑎𝑠𝑠/𝑒𝑛𝑒𝑟𝑔𝑦 𝑒𝑛𝑡𝑒𝑟𝑖𝑛𝑔 𝑎 𝑠𝑦𝑠𝑡𝑒𝑚 = 𝑅𝑎𝑡𝑒 𝑜𝑓 𝑚𝑎𝑠𝑠/𝑒𝑛𝑒𝑟𝑔𝑦 𝑙𝑒𝑎𝑣𝑖𝑛𝑔 𝑎 𝑠𝑦𝑠𝑡𝑒𝑚

Depending on a given process, the relevant mass and/or energy balance can be used.
The above equations look simple, but in reality there may be many in and out terms,
particularly for component balances. Also note that the mass accumulation in a system
𝑑𝑚
has the form , where m is the total mass in the system. Similarly, the rate of energy
𝑑𝑡
𝑑𝐸
accumulation has the form , where E is the total energy in a system. The equation
𝑑𝑡

can also be used for mole balances, and in that case, n will represent moles. The
𝑑𝑛
accumulation term will be represented by .
𝑑𝑡

In mathematical terms, the equations become:

𝑑𝑚
Mass balance: = 𝑚̇𝑖𝑛 − 𝑚̇𝑜𝑢𝑡 + 𝑚̇𝑔𝑒𝑛 − 𝑚̇𝑟𝑒𝑎𝑐
𝑑𝑡
𝑑𝐸
Energy balance: = 𝐸̇𝑖𝑛 − 𝐸̇𝑜𝑢𝑡 + 𝐸̇𝑔𝑒𝑛 − 𝐸̇𝑟𝑒𝑎𝑐
𝑑𝑡
𝑑𝑛
Mole balance: = 𝑛̇ 𝑖𝑛 − 𝑛̇ 𝑜𝑢𝑡 + 𝑛̇ 𝑔𝑒𝑛 − 𝑛̇ 𝑟𝑒𝑎𝑐
𝑑𝑡

16
PRC3701 LEARNING UNIT 2
Bearing in mind that 𝑚 = 𝜌𝑉, the mass balance can be written as:
𝑑(𝜌𝑉)
Mass balance: = 𝜌𝑖𝑛 𝐹𝑖𝑛 − 𝜌𝑜𝑢𝑡 𝐹𝑜𝑢𝑡 + 𝜌𝑔𝑒𝑛 𝐹𝑔𝑒𝑛 − 𝜌𝑟𝑒𝑎𝑐 𝐹𝑟𝑒𝑎𝑐
𝑑𝑡

This is usually applicable to fluids.

If we are dealing with gases which can be assumed to be ideal, then 𝑃𝑉 = 𝑛𝑅𝑇 can be
used to interchange the relationship to fit the given data. We could have a differential
equation in terms of pressure or even temperature, if required.

2.3. MODELLING OF A SIMPLE EXAMPLE

The principles you have learnt, when combined, allow us to devise a model of a
process. As illustration we will use a simple example of filling a tank with water and
controlling the level.

Example 2.1: Filling a tank with water and controlling the level
Water flows into a tank at the rate F in and is also permitted to flow out at a different
flowrate, F. The cross-sectional area of the tank is considered uniform, with a constant
value A.
(a) There are control valves at both the inlet and outlet pipes to assist in controlling
inlet and outlet flows.
(b) There is a differential pressure level measuring device to monitor and measure
water level.
(c) There are flow measurement devices for monitoring both inlet flow and outlet
flow of water.

All this is summarised in Figure 2.1.

17
PRC3701 LEARNING UNIT 2
Figure 2.1: Filling a tank with water

We want to design a control system that will maintain a constant water level in the tank.
The inlet and outlet flow valves are allowed to vary around their nominal values. An
electronic controller is available.

Solution:
Using the four principles you have learnt:
Step 1 – Assess the process and define the control objective(s).
We need to maintain a constant water level.

Step 2 – Identify the process (input and output) variables that will be used to achieve the
control objective.
Process variables:
Input variables: Fin – it is a disturbance variable
F – it is a manipulated variable
Output variable: h – it is a controlled variable

Step 3 – Decide on the control structure to use.

We will use feedback control. We will measure the level of water in the tank and
either increase or decrease F to maintain h, at the set point.

Step 4 – Design the controller. Here we need to obtain a model that relates inputs to

18
PRC3701 LEARNING UNIT 2
outputs using historical information and see whether it is able to predict what may happen
when there are changes. When we are developing a model, we need to consider two
aspects. These are:
Assumptions:
Steady-state conditions
Water density is constant, 

Model:
Mass balance:
𝐴𝑐𝑐 = 𝑚𝑎𝑠𝑠 𝑖𝑛 − 𝑚𝑎𝑠𝑠 𝑜𝑢𝑡 + 𝑚𝑎𝑠𝑠 𝑔𝑒𝑛𝑒𝑟𝑎𝑡𝑒𝑑 − 𝑚𝑎𝑠𝑠 𝑟𝑒𝑎𝑐𝑡𝑒𝑑
If there are no chemical reactions, there is no mass generated or mass reacted.
Therefore:
𝐴𝑐𝑐 = 𝑚𝑎𝑠𝑠 𝑖𝑛 − 𝑚𝑎𝑠𝑠 𝑜𝑢𝑡
𝑑𝑚
= 𝑚𝑎𝑠𝑠𝑖𝑛 − 𝑚𝑎𝑠𝑠𝑜𝑢𝑡
𝑑𝑡
𝑑(𝜌𝑉)
= 𝜌𝐹𝑖𝑛 − 𝜌𝐹
𝑑𝑡

But 𝜌𝑉 = 𝜌𝐴ℎ, so
𝑑ℎ
𝜌𝐴 = 𝜌𝐹𝑖𝑛 − 𝜌𝐹, density and cross-sectional area are constants. Simplifying
𝑑𝑡

further:
𝑑ℎ
𝐴 = 𝐹𝑖𝑛 − 𝐹 eqn 2.1
𝑑𝑡

At steady state, the time derivative vanishes, and if the inlet flow and outlet flow
have respective steady-state values of Fi,s and Fs, then eqn 2.1 becomes:
0 = 𝐹𝑖,𝑠 − 𝐹𝑠 eqn 2.2
Subtracting eqn 2.2 from eqn 2.1 gives:
𝑑ℎ
𝐴 = (𝐹𝑖𝑛 − 𝐹𝑖,𝑠 ) − (𝐹 − 𝐹𝑠 )
𝑑𝑡

This is the mathematical model of the above process.

Response of model
Let us consider how this system will respond to the situation in which the inlet
flowrate Fin suddenly increases from its steady-state value of Fi,s to (Fi,s + f). This
means that Fin becomes a disturbance. The simplest feedback control law we can
19
PRC3701 LEARNING UNIT 2
 use is proportional control (which you will learn later). Mathematically, the
relationship is as follows:
𝐹 − 𝐹𝑠 ∝ ℎ − ℎ𝑠 , or
𝐹 = 𝐹𝑠 + 𝐾(ℎ − ℎ𝑠 )
Where K is the proportionality constant

Using the above control law, an increase in the value of h over its steady-state value hs
will mean an increase in outlet flowrate F, so as to bring down the water level. Observe,
therefore, that this controller appears to make sensible control decisions.

DOF analysis
From the mathematical model:
Number of variables = 7 {A; h; t; Fin; Fi,s; F; Fs}
Number of equations = 1

But number of variables can be reduced to unknown only:

Known variables = A (specify); Fin; Fi,s; Fs
Unknown variables = h; t; F

So, because we have 3 unknown variables, we will need to formulate 2 additional

equations to solve the model.

Changing configurations
In this example, step 2 was specified as:
Process variables:
Input variables: Fin – it is a disturbance variable
F – it is the manipulated variable
Output variable: h – it is the controlled variable
And step 3 as feedback control.

This configuration can be changed, giving rise to different calculations as well. Here
are other configurations that can be used:
20
PRC3701 LEARNING UNIT 2
Configuration 2:
Step 2:
Process variables:
Input variables: Fin – it is a disturbance variable
F – it is the manipulated variable
Output variable: h – it is the controlled variable
Step 3: Feedforward control

Configuration 3:
Step 2:
Process variables:
Input variables: Fin – it is no longer a disturbance variable
Fin – it is the manipulated variable
Output variable: h – it is the controlled variable
Step 3: Feedback control
With a change in configuration, the model will also change. For configuration 3, the
model becomes:
𝑑ℎ
𝐴 = 𝐹𝑖𝑛 − 𝑐ℎ
𝑑𝑡

Where c is the constant of proportionality accounting for the valve resistance.

The process model equations we have dealt with so far have been linear, and so they
have been easy to work with. However, this is not always the case. In reality, outlet
flowrate is often proportional to the square root of liquid level, implying that a mor e
realistic model will be:
𝑑ℎ
𝐴 = 𝐹𝑖𝑛 − 𝑐√ℎ
𝑑𝑡

This is a non-linear equation. Modelling this process control system will be difficult, as
usually we require linear systems.

21
PRC3701 LEARNING UNIT 2
 2.4. LINEARISATION OF NON-LINEAR SYSTEMS
The theory of process control is based almost exclusively on linear systems. Thus far
we have been fortunate to have differential equations that are linear. In real life most
of the differential equations are non-linear, and so they need to be linearised before
being used as models. We will now turn our attention to the technique of linearisation
which is based on the Taylor series.

But before we do, let’s make sure we are certain about what is meant by linear
equations and linear systems. In general, “linear” refers to something that can be
represented by or is a straight line. A linear equation is simply a first-order polynomial,
and a linear differential equation is the sum of linear terms.

What, then, is linearisation? It is the process by which we approximate non-linear

systems with linear ones. This is an approximation term of the original term, but in a
linear form. This is to enable our models to be useful in process control.

2.4.1 Linearisation using Taylor series

Recap: Taylor series expansion
Consider the function f(x). If we want to use Taylor series expansion about a point x =
a, the expansion becomes:
𝑓′′ (𝑎) 𝑓′′′ (𝑎) 𝑓𝑛 (𝑎)
𝑓(𝑥) = 𝑓(𝑎) + 𝑓 ′ (𝑎)(𝑥 − 𝑎) + (𝑥 − 𝑎)2 + (𝑥 − 𝑎)3 + ⋯ + (𝑥 − 𝑎)𝑛 + ⋯
2! 3! 𝑛!

Where f’, f’’, f’’’, …, f n are first, second, third ,…, nth differentiation terms about the
value a. Usually only the first two to three terms are considered to approximate any
function.

Now consider the following non-linear differential equation, modelling a given chemical
engineering process:
𝑑𝑥
= 𝑓(𝑥) eqn 2.3
𝑑𝑡

We can expand this non-linear function f(x) into a Taylor series around the point x o.
The function becomes:

22
PRC3701 LEARNING UNIT 2
 𝑓′′ (𝑥𝑜 ) 𝑓′′′ (𝑥𝑜 ) 𝑓𝑛 (𝑥𝑜 )
𝑓(𝑥) = 𝑓(𝑥𝑜 ) + 𝑓 ′ (𝑥𝑜 )(𝑥 − 𝑥𝑜 ) + (𝑥 − 𝑥𝑜 )2 + (𝑥 − 𝑥𝑜 )3 + ⋯ + (𝑥 −
2! 3! 𝑛!

𝑥𝑜 )𝑛 + ⋯ eqn 2.4

Alternatively, in some books this is written as:

𝑑𝑓 𝑑2𝑓 (𝑥−𝑥𝑜 )2 𝑑3 𝑓 (𝑥−𝑥𝑜 )3 𝑑𝑛𝑓 (𝑥−𝑥𝑜 )𝑛
𝑓(𝑥) = 𝑓(𝑥𝑜 ) + ( ) (𝑥 − 𝑥𝑜 ) + ( 2) +( 3) + ⋯+ ( 𝑛) +
𝑑𝑥 𝑥𝑜 𝑑𝑥 𝑥𝑜 2! 𝑑𝑥 𝑥𝑜 3! 𝑑𝑥 𝑥𝑜 𝑛!

⋯ eqn 2.4

If we neglect all terms of order two and higher, we take the following approximation for
the value of f(x):
𝑓(𝑥) = 𝑓(𝑥𝑜 ) + 𝑓 ′ (𝑥𝑜 )(𝑥 − 𝑥𝑜 ) eqn 2.5

Alternatively, in some books this is written as:

𝑑𝑓
𝑓(𝑥) = 𝑓(𝑥𝑜 ) + ( ) (𝑥 − 𝑥𝑜 ) eqn 2.5
𝑑𝑥 𝑥𝑜

This can be shown on a graph as:

Figure 2.2: Linear approximation of a non-linear function

In eqn 2.3, replace f(x) by its linear approximation given by eqn 2.5, to get:

23
PRC3701 LEARNING UNIT 2
𝑑𝑥
= 𝑓(𝑥𝑜 ) + 𝑓 ′ (𝑥𝑜 )(𝑥 − 𝑥𝑜 ) eqn 2.6
𝑑𝑡

Example 2.2
Consider the tank system shown in Figure 2.3.

Figure 2.3: Tank system

The total mass balance yields:

𝑑ℎ
𝐴 = 𝐹𝑖 − 𝐹𝑜
𝑑𝑡

Where A is the cross-sectional area of the tank and h the height of the liquid level.

If the outlet flowrate F is a linear function of the level of the liquid, then:
𝐹𝑜 = 𝛼ℎ
where  is a constant

Then:
𝑑ℎ 𝑑ℎ
𝐴 = 𝐹𝑖 − 𝛼ℎ or 𝐴 + 𝛼ℎ = 𝐹𝑖
𝑑𝑡 𝑑𝑡

This is a linear differential equation, and no approximation is needed.

However, if 𝐹𝑜 = 𝛽√ℎ, then:

24
PRC3701 LEARNING UNIT 2
 𝑑ℎ 𝑑ℎ
𝐴 = 𝐹𝑖 − 𝛽√ℎ or 𝐴 + 𝛽√ℎ = 𝐹𝑖
𝑑𝑡 𝑑𝑡

This is a non-linear differential equation, and approximation will be required. Let us

develop the linearised approximation for this non-linear model. The only non-linear
term in the equation is β√h. So we take the Taylor series expansion of this term around
a point ho:
𝑑{𝛽√ℎ} 𝑑 2 {𝛽√ℎ} (ℎ−ℎ𝑜 )2
𝛽√ℎ = 𝛽√ℎ𝑜 + ( ) (ℎ − ℎ𝑜 ) + ( ) +⋯
𝑑ℎ ℎ=ℎ𝑜 𝑑ℎ2 ℎ=ℎ𝑜 2!

𝛽 𝛽
𝛽√ℎ = 𝛽√ℎ𝑜 + (ℎ − ℎ𝑜 ) − (ℎ − ℎ𝑜 )2 + ⋯
2√ℎ𝑜
83 √ℎ𝑜2

Neglecting terms of order two and higher, we have:

𝛽
𝛽√ℎ ≈ 𝛽√ℎ𝑜 + (ℎ − ℎ𝑜 )
2√ℎ𝑜

If introduced in the non-linear dynamic system, the equation yields the following
linearised approximate model:
𝑑ℎ 𝛽
𝐴 + 𝛽√ℎ𝑜 + (ℎ − ℎ𝑜 ) = 𝐹𝑖
𝑑𝑡 2√ℎ𝑜

Expanding and rearranging gives:

𝑑ℎ 𝛽 𝛽
𝐴 + ℎ = 𝐹𝑖 − √ℎ𝑜
𝑑𝑡 2√ℎ𝑜 2

The linearised model initially approximates the non-linear model very well. As time
goes on, however, the linearised approximation becomes progressively less accurate.

2.4.2 Deviation variables

Suppose that xs is the steady-state value of x describing the initial dynamic system, eqn
2.3:
𝑑𝑥𝑠
= 0 = 𝑓(𝑥𝑠 ) eqn 2.7
𝑑𝑡

Consider xs the point of linearisation for eqn 2.3. Then:

𝑑𝑥
= 𝑓(𝑥𝑠 ) + 𝑓 ′ (𝑥𝑠 )(𝑥 − 𝑥𝑠 ) eqn 2.8
𝑑𝑡

25
PRC3701 LEARNING UNIT 2
Subtracting eqn 2.7 from eqn 2.8 gives:
𝑑(𝑥−𝑥𝑠 )
= 𝑓 ′ (𝑥𝑠 )(𝑥 − 𝑥𝑠 ) eqn 2.9
𝑑𝑡

We can then define a deviation variable x’ as:

𝑥 ′ = 𝑥 − 𝑥𝑠

Therefore, eqn 2.9 becomes:

𝑑𝑥 ′
= 𝑓 ′ (𝑥 ′ ) eqn 2.10
𝑑𝑡

Equation 2.10 is the linearised approximation of the non-linear dynamic system, equation
2.3, expressed in terms of the deviation variable x’.

IN SUBSEQUENT UNITS WE WILL MAKE USE OF DIFFERENTIAL EQUATIONS, IN

TERMS OF DEVIATION VARIABLES.

Example 2.3
𝑑ℎ
Consider the linearised model of the tank system (𝐴 + 𝛽√ℎ = 𝐹𝑖 ). Let hs be the steady-
𝑑𝑡
state value of the liquid level for a given value, Fi,s of the inlet flow rate. Then the
linearised model around hs gives:
𝑑ℎ 𝛽 𝛽
𝐴 + ℎ = 𝐹𝑖 − √ℎ𝑠 eqn 2.11
𝑑𝑡 2√ℎ𝑠 2

𝑑ℎ
At steady state 𝐴 + 𝛽√ℎ = 𝐹𝑖 becomes:
𝑑𝑡

𝑑ℎ𝑠
𝐴 + 𝛽√ℎ𝑠 = 𝐹𝑖,𝑠 eqn 2.12
𝑑𝑡

26
PRC3701 LEARNING UNIT 2
Subtract eqn 2.12 from eqn 2.11:
𝑑(ℎ−ℎ𝑠 ) 𝛽
𝐴 + (ℎ − ℎ𝑠 ) = 𝐹𝑖 − 𝐹𝑖,𝑠 eqn 2.13
𝑑𝑡 2√ℎ𝑠

Defining the deviation variables:

ℎ′ = ℎ − ℎ𝑠 and 𝐹𝑖′ = 𝐹𝑖 − 𝐹𝑖,𝑠

and the linearised form in terms of deviation variables is:

𝑑ℎ′ 𝛽
𝐴 + ℎ′ = 𝐹𝑖′ eqn 2.14
𝑑𝑡 2√ℎ𝑠

2.4.3 Linearisation of systems with many variables

In process control we often encounter systems with more than one variable. Consider
the following dynamic system:
𝑑𝑥1
= 𝑓1 (𝑥1 , 𝑥2 ) eqn 2.15
𝑑𝑡

𝑑𝑥2
= 𝑓2 (𝑥1 , 𝑥2 ) eqn 2.16
𝑑𝑡

By expanding the non-linear functions f1(x1,x2) and f2(x1,x2) into Taylor series around the
point (x1,0; x2,0) and neglecting the terms of order two and higher, we have:
𝑑𝑥1 𝑑𝑓1 𝑑𝑓1
= 𝑓1 (𝑥1,0 ; 𝑥2,0 ) + [ ] (𝑥1 − 𝑥1,0 ) + [ ] (𝑥2 − 𝑥2,0 ) eqn 2.17
𝑑𝑡 𝑑𝑥1 (𝑥 ;𝑥 ) 𝑑𝑥2 (𝑥 ;𝑥 )
1,0 2,0 1,0 2,0

𝑑𝑥2 𝑑𝑓2 𝑑𝑓2

= 𝑓2 (𝑥1,0 ; 𝑥2,0 ) + [ ] (𝑥1 − 𝑥1,0 ) + [ ] (𝑥2 − 𝑥2,0 ) eqn 2.18
𝑑𝑡 𝑑𝑥1 (𝑥 ;𝑥 ) 𝑑𝑥2 (𝑥 ;𝑥 )
1,0 2,0 1,0 2,0

Equations 2.17 and 2.18 are linear differential equations and represent the linearised,
approximate model of the initial non-linear system described by equations 2.15 and 2.16.

27
PRC3701 LEARNING UNIT 2
 2.5. CONCLUSION
This learning unit has given you a basic introduction to how we formulate mathematical
models of chemical engineering processes which we can use as models to represent
the processes. These models are useful in process control, and they are usually
formulated as differential equations. Process control works well with linear differential
equations, and so if the model is non-linear, the Taylor series expansion up to second
term can be used to approximate the non-linear system. The principles will be extremely
useful in the units to follow. Taylor series expansion is always about a steady-state point
which can introduce deviation variables.

Thus far we have formulated the model in differential equation form, but we have not yet
solved the model to determine the relationship between the input and output variables.
The next learning unit introduces a concept that makes it possible to solve the differential
equations in half the time we would take using other mathematical principles. This
concept is integration.

28
PRC3701 LEARNING UNIT 2
 2.6. PROBLEMS

You now have the opportunity to apply what you have learnt in this unit. Please solve
the following problems.

Problem 2.1

Linearise the following function about Ts

𝑇
𝑓(𝑇) = 𝐴 sin ( )
2𝜋

Problem 2.2

Discuss the following problem in Discussions on the module site on myUnisa (Forum 5:
Learning unit 2 problems, Problem 2.2).
In the tank problem (Example 2.2), if the inlet flowrate were decreased from the steady-
state value and the two solutions allowed to approach their respective steady-state
heights, would the analytical height be more or less than that predicted by the linear
solution? Explain your reasoning.

Share your answer in Discussions. Also remember to read the contributions made by
your fellow students.

Problem 2.3

What are the deviation variables? What is the point of linearisation in order to define the
deviation variables that will be useful for process control purposes?

Problem 2.4

Controlling operation of stirred tank heater:

Consider the tank heater system shown in Figure 2.4. A liquid enters the tank at a
flowrate Fi (m3/s) and a temperature Ti (oC), where it is heated with steam (having a
flowrate of Fst, kg/s). Let F and T be the flowrate and temperature of the steam leaving
the tank. The tank is considered to be well stirred, which means that the temperature of
29
PRC3701 LEARNING UNIT 2
the effluent is equal to the temperature of the liquid in the tank. The operational
objectives of this heater are:

o to keep the effluent temperature T at the desired value Ts

o to keep the volume of the liquid in the tank at the desired value Vs

Use the principles you have learnt to devise the model(s) for this process. Identify all
the required parameters as per the four steps of modelling principles.

Figure 2.4: Stirred tank heater system

Problem 2.5

Linearise the non-linear equation:

𝑧 = 𝑥𝑦

in the region 5  x  7, 10  y  12. Find the error if the linearised equation is used to
calculate the value of z when x = 5, y = 10.

Problem 2.6

Linearise the non-linear equation:

𝑧 = 𝑥 2 + 4𝑥𝑦 + 6𝑦 2

30
PRC3701 LEARNING UNIT 2
in the region defined by 8  x  10, 2  y  4. Find the error if the linearised equation is
used to calculate the value of z when x = 8, y = 2.

31
PRC3701 LEARNING UNIT 2
 2.7. SOLUTIONS TO PROBLEMS

Compare your solutions with the solutions provided below.

Solution 2.1
𝑇
The linearised form of the function 𝐴 sin ( ) is:
2𝜋

𝑇 𝑇−𝑇𝑠 𝑇−𝑇𝑠
𝐴 sin ( 𝑠 ) + [ ] 𝐴 cos ( )
2𝜋 2𝜋 2𝜋

Solution 2.2

You will find the solution to this problem on the module site on myUnisa, Forum 5:
Learning unit 2, Solution 2.2.

Solution 2.3

The deviation variable is the variation of a process from its steady state, which must be
corrected by some process control.

Thus: if xs describes the steady state, then 𝑥 − 𝑥𝑠 = 𝑥 ′ is the deviation variable.

Linearising around the steady-state point leads one to a linearised approximation of a

non-linear system.

Solution 2.4

The process is as follows:

Step 1 – Assess the process and define the control objective(s)

To keep the effluent temperature T at the desired value Ts

To keep the volume V of the liquid in the tank at the desired value Vs

32
PRC3701 LEARNING UNIT 2
Step 2 – Identify the process (input and output) variables that will be used to achieve the
control objective. When carrying out this step we need to ask ourselves questions like:
Process variables:
Input variables
Disturbances: Fi and Ti
Manipulated variables: Fst

Output variables
Controlled variables T and V
Manipulated variables: F

Step 3 – Decide on the control structure to use

We will use feedback control. We will measure the volume of the liquid and the
outlet temperature of the liquid and either increase or decrease Fi and Fst to
maintain V and T, at the set points of Vs and Ts.

Step 4 – Design the controller. Here we need to obtain a model that relates inputs to
outputs using historical information and see whether it is able to predict what may happen
when there are changes. When we are developing a model, we need to consider two
aspects. These are:
Assumptions:
Steady-state conditions
The tank is considered to be well stirred, which means that the temperature of the
effluent is equal to the temperature of the liquid in the tank.
Liquid density is constant, 

The two models (one from mass balances and one from energy balances):
Mass balance:
𝐴𝑐𝑐 = 𝑚𝑎𝑠𝑠 𝑖𝑛 − 𝑚𝑎𝑠𝑠 𝑜𝑢𝑡
𝑑𝑚
= 𝑚𝑎𝑠𝑠𝑖𝑛 − 𝑚𝑎𝑠𝑠𝑜𝑢𝑡
𝑑𝑡

33
PRC3701 LEARNING UNIT 2
 𝑑(𝜌𝑉)
= 𝜌𝐹𝑖 − 𝜌𝐹
𝑑𝑡

Density is constant, so:

𝑑𝑉
𝜌 = 𝜌𝐹𝑖 − 𝜌𝐹
𝑑𝑡

Simplifying further:
𝑑𝑉
= 𝐹𝑖 − 𝐹 eqn 1
𝑑𝑡

At steady state, the time derivative vanishes, and if the inlet flow and outlet flow
have respective steady-state values of Fi,s and Fs, then eqn 1 becomes:
0 = 𝐹𝑖,𝑠 − 𝐹𝑠 eqn 2
Subtracting eqn 2 from eqn 1 gives:
𝑑𝑉
= (𝐹𝑖 − 𝐹𝑖,𝑠 ) − (𝐹 − 𝐹𝑠 ) eqn 3
𝑑𝑡

This is the mathematical model of the above process in terms of controlling the
liquid volume.

Energy balance:
𝐴𝑐𝑐 = 𝑒𝑛𝑒𝑟𝑔𝑦 𝑖𝑛 − 𝑒𝑛𝑒𝑟𝑔𝑦 𝑜𝑢𝑡
𝑑𝑇
𝑉𝜌𝐶𝑝 = {𝐹𝜌𝐶𝑝 𝑇𝑖 + 𝑄} − 𝐹𝜌𝐶𝑝 𝑇 eqn 4
𝑑𝑡

Simplifying further:
𝑑𝑇
𝑉𝜌𝐶𝑝 = 𝐹𝜌𝐶𝑝 (𝑇𝑖 − 𝑇) + 𝑄 eqn 4
𝑑𝑡

At steady state there is no accumulation:

0 = 𝐹𝜌𝐶𝑝 (𝑇𝑖,𝑠 − 𝑇𝑠 ) + 𝑄𝑠 eqn 5

Subtracting eqn 5 from eqn 4:

𝑑𝑇
𝑉𝜌𝐶𝑝 = 𝐹𝜌𝐶𝑝 [(𝑇𝑖 − 𝑇𝑖,𝑠 ) − (𝑇 − 𝑇𝑠 )] + (𝑄 − 𝑄𝑠 ) eqn 6
𝑑𝑡

This is the mathematical model of the above process.

Response of model
Let us consider how this system will respond to the situation in which the inlet
flowrate Ti suddenly increases from its steady-state value of Tis to (Tis + t). This

34
PRC3701 LEARNING UNIT 2
 means that Ti becomes a disturbance. The simplest feedback control law we can
use is proportional control (which you will learn about later).
Mathematically, the relationship for change in heat Q is as follows:
𝑄 − 𝑄𝑠 ∝ −(𝑇 − 𝑇𝑠 ) , or 𝑄 = 𝑄𝑠 − 𝛼(𝑇 − 𝑇𝑠 )
where  is proportionality constant and the negative sign is to show the inverse
relationship between heat, Q and outlet temperature, T.

Using the above control law, an increase in the value of T over its steady-state
value Ts will mean a decrease in steam outlet flowrate Q, so as to bring down the
outlet temperature. This controller therefore appears to make sensible control
decisions.

Substituting for Q gives:

𝑑𝑇
𝑉𝜌𝐶𝑝 = 𝐹𝜌𝐶𝑝 [(𝑇𝑖 − 𝑇𝑖,𝑠 ) − (𝑇 − 𝑇𝑠 )] − 𝛼(𝑇 − 𝑇𝑠 ) eqn 5
𝑑𝑡

Solution 2.5

Since the region considered is given by 5  x  7, 10  y  12, choose 𝑥̅ = 6, and 𝑦̅ =

11. Then:

𝑧̅ = 𝑥̅ 𝑦̅ = 66

Let us obtain a linearized equation for the non-linear equation near a point 𝑥̅ = 6, 𝑦̅ =
11. Expanding the non-linear equation into a Taylor series about a point and neglecting
the higher-order terms, we have:

𝑓(𝑥) = 𝑓(𝑎) + 𝑓 ′ (𝑎)(𝑥 − 𝑎)

𝑓(𝑥) − 𝑓(𝑎) = 𝑓 ′ (𝑎)(𝑥 − 𝑎)

𝜕𝑥 𝜕𝑦
𝑧 − 𝑧̅ = (𝑥𝑦){𝑥 − 𝑥̅ } + (𝑥𝑦){𝑦 − 𝑦̅}
𝜕𝑡 𝑦̅=11 𝜕𝑡 𝑥̅ =6

𝑧 − 𝑧̅ = 𝑦̅{𝑥 − 𝑥̅ } + 𝑥̅ {𝑦 − 𝑦̅}

Substituting terms:

𝑧 − 66 = 11{𝑥 − 6} + 6{𝑦 − 11}

Simplifying:
35
PRC3701 LEARNING UNIT 2
𝑧 = 11𝑥 + 6𝑦 − 66

When x = 5, y = 10, the value of z given by the linearised equation is:

𝑧 = 11𝑥 + 6𝑦 − 66 = 11(5) + 6(10) − 66 = 49

The exact value of z is:

𝑧 = 𝑥𝑦 = 5(10) = 50

The error is thus:

𝑒𝑟𝑟𝑜𝑟 = 50 − 49 = 1
50−49
%𝑒𝑟𝑟𝑜𝑟 = × 100% = 2%
50

Solution 2.6

Since the region considered is given by 8  x  10, 2  y  4, choose 𝑥̅ = 9, and 𝑦̅ = 3.

Then:

𝑧̅ = 𝑥̅ 2 + 4𝑥̅ 𝑦̅ + 6𝑦̅ 2 = 92 + 4(9)(3) + 6(32 ) = 243

Let us obtain a linearized equation for the non-linear equation near a point 𝑥̅ = 9, 𝑦̅ = 3.
Expanding the non-linear equation into a Taylor series about a point and neglecting the
higher order terms, we have:

𝑓(𝑥) = 𝑓(𝑎) + 𝑓 ′ (𝑎)(𝑥 − 𝑎)

𝑓(𝑥) − 𝑓(𝑎) = 𝑓 ′ (𝑎)(𝑥 − 𝑎)

𝜕𝑥 𝜕𝑦
𝑧 − 𝑧̅ = (𝑥 2 + 4𝑥𝑦 + 6𝑦 2 ){𝑥 − 𝑥̅ } + (𝑥 2 + 4𝑥𝑦 + 6𝑦 2 ){𝑦 − 𝑦̅}
𝜕𝑡 𝑦̅=3 𝜕𝑡 𝑥̅ =9

𝑧 − 𝑧̅ = (2𝑥̅ + 4𝑦̅){𝑥 − 𝑥̅ } + (4𝑥̅ + 12𝑦̅){𝑦 − 𝑦̅}

Substituting terms:

𝑧 − 243 = [2(9) + 4(3)]{𝑥 − 9} + [4(9) + 12(3)]{𝑦 − 3}

Simplifying:

36
PRC3701 LEARNING UNIT 2
𝑧 = 30(𝑥 − 9) + 72(𝑦 − 3) + 243

𝑧 = 30𝑥 − 270 + 72𝑦 − 216 + 243

𝑧 = 30𝑥 + 72𝑦 − 243

When x = 8, y = 2, the value of z given by the linearised equation is:

𝑧 = 30𝑥 + 72𝑦 − 243 = 30(8) + 72(2) − 243 = 141

The exact value of z is:

𝑧 = 𝑥 2 + 4𝑥𝑦 + 6𝑦 2 = 82 + 4(8)(2) + 6(22 ) = 152

The error is thus:

𝑒𝑟𝑟𝑜𝑟 = 152 − 141 = 11

152−141
%𝑒𝑟𝑟𝑜𝑟 = × 100% = 7.2%
152

37
PRC3701 LEARNING UNIT 2