Modelling of Control Systems
Dr. ASHRAF E. SUYYAGH
THE UNIVERSITY OF JORDAN
DEPARTMENT OF COMPUTER ENGINEERING
SPRING 2023
Introduction to System Modelling 2
Modeling refers to the process of describing the nature, systems and subsystems around us.
We usually model systems by observing them, then coming up with mathematical equations that describe the systems and their
behavior. We observe the outputs across a range of inputs, and we can use for example curve fitting to find the best equation
that represents the system with least errors. A good model is one that reflects the actual system behavior accurately. The outputs
for certain inputs in our model should match the actual system in real-life.
Sometimes it is difficult to get an exact model due to system complexity, but an a very good approximation can be good enough
in many cases. Engineering is the art of approximation!
In control systems, modelling requires applying the fundamental physical laws of science and engineering: For example, Ohm’s
law, Kirchhoff’s laws, Newton’s laws, laws of motion or heat transfer, etc. Some of these equations are linear equations, others
are non-linear.
Many of these governing equations have derivatives, integrals, and many are described in differential equations. Other systems
might have dynamical models, statistical models, or gam theoretical models. We concern ourselves in this course with linear
systems. The definition of linearity depends on context. Yet, in general most of nature is not always linear. We deal with
approximations to simplify the design, For example, many systems can be treated as linear within a certain range. Or we can
use step-wise approximation to model the system no-linear parts as approximated linear steps. Non-linear systems are handled
in advanced control systems courses.
Quick Review of Linear Systems 3
Linear systems must satisfy two principles:
The Principle of Superposition (Additivity)
In general, a necessary condition for a linear system can be determined in terms of an excitation (input) x(t) and a response (output) y(t).
▪ When the system at rest is subjected to an excitation x1(t), it provides a response y1(t).
▪ Furthermore, when the system is subjected to an excitation x2(t), it provides a corresponding response y2(t).
▪ For a linear system, it is necessary that the excitation x1(t)+ x2(t) results in a response y1(t)+ y2(t).
Thie Property of Homogeneity
The magnitude scale factor must be preserved in a linear system. That is, if a system generates an output y(t) for an input x(t), it must produce
an output a*y(t) for an input a*x(t).
Most systems we will be dealing with are based on differential equations. These differential equations must also satisfy the above two principles
to be considered linear. But since you have not taken the related differential course, we will restrict the examples in the next slide to simpler non-
differential equations.
Linear and Non-Linear Systems Examples 4
Example 1: Y = 5X Example 2: y = 5eX
❖X1 = 3 → Y1 = 15 ❖X1 = 3 → Y1 = 100.4277
❖X2 = 4 → Y2 = 20 ❖X2 = 4 → Y2 = 272.9908
❖X1 + X2 = 7 → Y3 = 35 which is equal to 15 + 20 ❖X1 + X2 = 7 → Y3 = 5483.165792 ≠ 100.4277 + 272.9908
Check 1: The system satisfies the superposition property The system does not satisfy the principle of superposition and is
therefore non-linear
❖X1 = 10 → Y1 = 50
❖X2 = 2x10 → Y2 = 2 x 50 = 100
Check 2: The system satisfies the homogeneity property Example 3: y = 5x + 4
Therefore, the system is linear ❖X1 = 10 → Y1 = 54
❖X2 = 2 x 10 → Y2 = 104
The result 104 is not twice 54 when the input doubled, therefore
the system is non-homogenous → non-linear system
What if the systems are non-linear? 5
We have already said that the majority of systems in nature are non-linear. However,
for the most part, these physical systems are linear for some range of the variables.
For example, let us take a look at the operation of a FET transistor:
▪ It has approximately linear operation (amplification) if the input Vg is between 0 – 0.4 Volts
▪ It exhibits non-linear output beyond an input Vg = 0.4 Volts → Goes towards saturation
Similarly, a mechanical spring system is linear if the pull is within a certain input
force, but as the pulling force increases beyond a certain value, the response is non-
linear (the spring will not bounce back).
Therefore, in many engineering applications, the linearity of many mechanical and
electrical elements can be assumed over a reasonably large range of the variables.
However, this is not usually the case for thermal and fluid elements, which are more
frequently nonlinear in character.
We can approximate the non-linear part of the system as linear but there will be of
course some model errors.
Quick Review of Time Variant and Invariant Systems 6
Any system has inputs and outputs.
A time invariant system is a system in which the output does not depend on the time at which the input arrives. That is; regardless of when the input
comes in, whether now or delayed; the output to this input remains always the same result.
Sometimes, we can tell if the system is invariant system if the equation has no dependance on time as A SEPARATE variable.
V = IR (Time-invariant)
V(t) = I(t) R (Also time invariant)
A time variant system is a system in which the output changes if the same input arrives at different times. That is; if the input x = 5 arrives at t = 0, it will
produce a different result if x = 5 arrives delayed by 10 seconds.
We can spot a time variant system if the equation has dependance on time as A SEPARATE variable.
Y = tX (Time-variant, notice that t is a separate variable)
y(t) = tx(t) (Time-variant, notice that t is a separate variable)
A time invariant system might not be easy to spot, for example let us revise the Ohm’s law more accurately by taking the effects of temperature on the
resistor:
V(t) = I(t) R(t) where R(t) changes due to the varying component temperature over time (Time-variant)
In the Signals and Systems Course, and in this Control course, we assume we are working with only Linear Time-Invariant Systems (LTI)
7
The following slides show you that many basic elements have equations that have
integrals and differentials. These equations are related to basic elements that you
have studied in Physics 101 and 102 course. You are not required to memorize
these equations. They are for reference only. They serve to show you how writing
system equations in the time domain could get complicated.
Simple Dynamics of Electrical Components 8
Simple Dynamics of Mechanical Components (I) 9
Simple Dynamics of Mechanical Components (II) 10
Electrical Example of System Modelling 11
Simple Parallel RLC Circuit
One may describe the electrical RLC circuit shown in the figure by utilizing Kirchhoff’s current law (KCL). We will obtain the following
integrodifferential equation:
As you can see, it is not straightforward to take v(t) as a common factor and one
must solve the system.
For more complex systems, the equations get even more complicated with
a second derivative and a first derivative for a simple amplifier circuit:
Mechanical Example of System Modelling 12
Automobile Suspension System
A very simplified version of the suspension system is shown. Assuming that the motion (displacement) xi at point P is the input to the
system and the vertical motion xo of the body is the output, then, the equation of motion for the system is:
Notice the second derivative and a first derivative
Modelling Complexity as No. of Subsystems 13
Grows
Real systems have many electrical, mechanical, hydraulic, and pneumatic subsystems. Putting all these subsystems all together and
writing their equations in the time domain will expose many complexities. Therefore, trying to solve the system in the time domain
(e.g., finding the relationship between the output and the input – what we will later call the transfer function) can be extremely difficult.
Ex.1: Shuttle Control Ex.2: Hybrid Car Control
Circumventing the Complex Dynamics of 14
Control Systems
Most systems even when constructed from simple
components will quickly turn to be more complex to
analyze in the time domain, as many of the basic
components are governed by integrodifferential
equations as we have seen in previous slides.
Mathematicians and engineers have found out
brilliant ways to circumvent this complexity and solve
this hard problem → Laplace transform to the
rescue.
We use the Laplace transform to switch from the
time-domain to the s-domain. The Laplace transform
turns integrodifferential equations into algebraic
ones which are easy to solve.
▪ Furthermore, it is easier to analyze the system in the s-domain and gain much needed insights on how it works
Laplace Transform 15
The Laplace transform mathematically resembles the Fourier transform:
But here, s is a complex number, so s is α + jw
Fourier Transform is a special case of Laplace, when α = 0, where there are no underlying exponentials in the signal.
Laplace is a more generic transform.
Laplace’s transform beauty lies in its ability to transform functions, even integral, differential into simple ALGEBRAIC notation.
Algebra is much easier to solve.
Notice that Laplace is a continuous time-domain transformation into the spatial (frequency domain)
In discrete time systems, where we have samples fed into digital controllers, we have the equivalent z-transform
MATLAB has functions for Fourier, Laplace, and z-transforms and their inverse transformations.
We solve the system in the s-domain (continuous) or z-domain (discrete). Once we find the solutions, we can use the inverse
transform to get back to the time domain.
Laplace Transform Table 16
Laplace Transform Table Example I 17
From the Laplace Table, we can find this relationship:
So, in our case, a = -5 and n = 1, substituting in the
Laplace transform, we get:
Laplace Transform Table Example II 18
𝑑 2 𝑥(𝑡) 𝑑𝑥(𝑡)
+4 + 5𝑥 𝑡 = 1, 𝑠𝑢𝑝𝑝𝑜𝑠𝑒 𝑡ℎ𝑎𝑡 𝑥 0 = 1, 𝑥 ′ 0 = −1
𝑑𝑡 𝑑𝑡
From the Laplace Table, we can find this relationship:
𝒔𝟐 𝑿 𝒔 − 𝒔 ∗ 𝒙 𝟎 − 𝒙′ 𝟎 = 𝒔𝟐 𝑿 𝒔 − 𝒔 ∗ 𝟏 − (−𝟏) = 𝒔𝟐 𝑿 𝒔 − 𝒔 + 𝟏
4 𝒔𝑿 𝒔 − 𝒙 𝟎 = 4 𝒔𝑿 𝒔 − 𝟏 = 𝟒𝒔𝑿 𝒔 − 𝟒
5 X(s)
𝟏
𝒔
Overall:
1
s2 X s − s + 1 + 4sX s − 4 − 5 X s =
s
1 s2 +3s+1
s2 X s + 4sX s − 5 X s = +s+3=
s s
2 s2 +3s+1 s2 +3s+1
X(s)(s + 4s – 5) = ➔ X(s) =
s s(s2 + 4s – 5)
Laplace Transform Table Example III 19
The Transfer Function 20
A control system has an input signal (reference input) r(t) which is going to
provide an output (a control signal) c(t) as you can see in the simple open-
loop control system (Figure (a)).
When a system is complex, and it has many electrical or mechanical
components, we divide it into subsystems. But still, we have an input signal
r(t) which is going to provide an output (a control signal) c(t) as you can see
in the simple open-loop control system (Figure (b)).
For example, in the adjacent parallel RLC circuit, the current I (r(t)) controls
the output voltage v(t)
The relationship between the output and the input is governed by the
equations of the system.
We are interested to know the relationship of the output to the input; that is:
𝑣(𝑡)
G(t) = which we call the transfer function.
𝑟(𝑡)
But in the time domain It is not easy to isolate v(t)
The Transfer Function 21
This is why we use the Laplace transform and express the Transfer Function as:
𝑣(𝑠)
𝐺 𝑠 =
𝑟(𝑠)
1. Use Algebra instead of integrodifferential equations
2. Spatial domain provides means to analyze the system in ways unavailable to us in
the time-domain. Exposes and allows using tools to determine transient response,
steady-state, stability.
Assuming no initial conditions, that is v(0) = v’(0) = 0, and using Laplace Tables, we Transfer Functions are valid when the system is
can rewrite the system equations as: LINEAR; that is, the governing equations are
𝑉(𝑠)
+ 𝑠𝐶𝑉(𝑠) +
𝑉(𝑠)
= 𝑅(𝑠) linear.
𝑅 𝑠𝐿
If not, we need to apply linearization to non-
1 1 linear components first; then retrieve the transfer
𝑉 𝑠 + 𝑠𝐶 + =𝑅 𝑠 function.
𝑅 𝑠𝐿
𝑉(𝑠) 1
G 𝑠 = = Notice that multiplying by s denotes a derivative,
𝑅(𝑠) 1 1
+ 𝑠𝐶 + so the s works as a differentiator block
𝑅 𝑠𝐿
𝑉(𝑠) 𝑠𝑅𝐿 While dividing by s denotes an integration, so
G 𝑠 = = 2 the s works as an integrator block
𝑅(𝑠) 𝑠 𝑅𝐿𝐶 + 𝑠𝐿 + 𝑅
The Transfer Function of Complex Systems 22
Most control systems are far more complex.
1. Divide them into subsystems
2. Retrieve the transfer function for each subsystems individually
3. Simplify using block diagram reduction techniques or by Maths
Transfer Function of Systems with Negative Feedback 23
▪ The input to this system is the signal R(s)
▪ Since this is a negative feedback system, the actual controlling
signal is the difference (error) between the input signal and the
feedback signal (the previous reading) which is C(s) H(s)
▪ Therefore, the difference (error) signal is E(s) = R(s) – C(s)H(s)
▪ The new output signal C(s) is the result of applying the control
(system) G(s) to the difference (error) signal G(s)(R(s) – C(s)H(s))
So:
𝐶 𝑠 = 𝐺(𝑠)(𝑅 𝑠 − 𝐶 𝑠 𝐻(𝑠))
𝐶 𝑠 = 𝐺 𝑠 𝑅 𝑠 − 𝐺 𝑠 𝐶 𝑠 𝐻(𝑠)
𝐶 𝑠 + 𝐺 𝑠 𝐶 𝑠 𝐻(𝑠) = 𝐺 𝑠 𝑅 𝑠
A positive feedback system will have the same
𝐶 𝑠 (1 + 𝐺 𝑠 𝐻 𝑠 ) = 𝐺 𝑠 𝑅 𝑠 transfer function, except that the sign will be
negative in the denominator
𝐶(𝑠) 𝐺(𝑠)
= 1+𝐺
𝑅(𝑠) 𝑠 𝐻 𝑠
Examples 24
Example 1) Determine the overall transfer function for the adjacent control system which
has a negative feedback loop with a transfer function of a gain equal 4 and a forward path
transfer function of 2/(s + 2).
The overall transfer function of the system is:
Example 2) Determine the overall transfer function for the adjacent control system which
has a positive feedback loop with a transfer function of a gain equal 4 and a forward path
transfer function of 2/(s + 2).
The overall transfer function of the system is:
Partial Fractions Form 25
Now that we have found transfer functions that relate the control system
outputs to its inputs, this transfer function can be analyzed to find the
control system parameters that we have discussed in Chapter 1: settling
time, overshoot, stability, etc.
But first, we need to mathematically transform the final transfer function
into the partial fractions form, from which it is even easier to use the
inverse Laplace transform to get back to the time-domain. Though it is
mathematically easy to do so, we shall do it in MATLAB in self-learning
material.
For example: thy function Y(s) below can be rewritten as:
Partial Fractions Form is عكس توحيد المقام والعودة إلى الكسور األصلية
If you are interested in working out partial fractions mathematically, refer to
the slides at the end of this chapter.
Simulink 26
Simulink® is a block diagram environment for multidomain simulation and Model-Based Design. It supports system-level design,
simulation, automatic code generation, and continuous test and verification of embedded systems. Simulink provides a graphical
editor, customizable block libraries, and solvers for modeling and simulating dynamic systems. It is integrated with MATLAB® ,
enabling you to incorporate MATLAB algorithms into models and export simulation results to MATLAB for further analysis.
Simulink Features:
• Graphical editor for building and managing hierarchical block diagrams
• Libraries of predefined blocks for modeling continuous-time and discrete-time systems
• Simulation engine with fixed-step and variable-step ODE solvers
• Scopes and data displays for viewing simulation results
• Project and data management tools for managing model files and data
• Model analysis tools for refining model architecture and increasing simulation speed
• MATLAB Function block for importing MATLAB algorithms into models
• Legacy Code Tool for importing C and C++ code into models
What is Simulink? Introductory Video 27
Simulink and Simscape 28
Simscape provides an environment for modeling and simulating physical systems spanning mechanical, electrical, hydraulic, and
other physical domains. It provides fundamental building blocks from these domains that you can assemble into models of
physical components, such as electric motors, inverting op-amps, hydraulic valves, and ratchet mechanisms. Because Simscape
components use physical connections, your models match the structure of the system you are developing.
Simscape models can be used to develop control systems and test system-level performance. You can extend the libraries using
the MATLAB® based Simscape language, which enables text-based authoring of physical modeling components, domains, and
libraries. You can parameterize your models using MATLAB variables and expressions, and design control systems for your
physical system in Simulink® . To deploy your models to other simulation environments, including hardware in-the-loop (HIL)
systems, Simscape supports C-code generation.
Simscape Key Features
• Single environment for modeling and simulating mechanical, electrical, hydraulic, thermal, and other multidomain physical
systems
• Libraries of physical modeling blocks and mathematical elements for developing custom components
• MATLAB based Simscape language, enabling text-based authoring of physical modeling components, domains, and libraries
• Physical units for parameters and variables, with all unit conversions handled automatically
• Ability to simulate models that include blocks from related physical modeling products without purchasing those products
• Support for C-code generation
What is Simscape? Introductory Video 29
Math Review: Partial Fractions Example 1 (Optional) 30
Determine the partial fractions of:
The partial fractions are of the form:
Then, for the partial fraction expression to equal the original fraction, we must have:
and consequently:
Pick values of s that will enable some of the terms involving constants to become zero and so enable other constants to be
determined:
Let s = -2 then
so B = -2
Let s = -1 then
so A = 3
Therefore:
Math Review: Partial Fractions Example 2 (Optional) 31
Determine the partial fractions of:
The partial fractions are of the form:
Then, for the partial fraction expression to equal the original fraction, we must have:
and consequently:
We start equating the same power terms on each side to determine A, B, and C
Equating s2 terms gives 0 = A. Equating s terms gives 3 = 2A + B and so B = 3. Equating the numeric terms gives 1 = A + 2B + C
and so C =-5. Thus:
Math Review: Partial Fractions Example 3 (Optional) 32
Determine the partial fractions of:
The partial fractions are of the form:
Then, for the partial fraction expression to equal the original fraction, we must have:
and consequently:
We start equating the same power terms on each side to determine A, B, and C
With S = -2 then -3 = 3C and so C = -1. Equating s2 terms gives 0 = A +C and so A= 1. Equating s terms gives 2 = 2A + B+ C and so
B = 1. Thus
Math Review: Partial Fractions Example 4 (Optional) 33
Determine the partial fractions of: Notice the power of numerator and denominator is the same
We must first use division:
And the expression becomes:
Then, for the partial fraction expression to equal the original fraction, we must have:
and consequently:
We start equating the same power terms on each side to determine A, B, and C
With s = 2, then B = 5/3. With s =-4, then A = -17/3. Thus, the expression can be written as:
References and Textbook Material 34
The material in these slides are based on:
Control Systems Engineering, Norman S. Nise, 7th Edition (2014), John Wiley And Sons
• Chapter 2 – System Modelling in the Frequency Domain
Sections 2.1, 2.2, 2.3, 2.4, 2.5, 2.6 (Students kindly note that these sections involve lots
of math, and we only described the ideas as we will use MATLAB instead)