EE

3114 Systems and Control

2. Mathematical Modeling

Semester A, 2019-20
Mathematical Representation of a System

To control a system, a mathematical representation

of the system is needed, which can be:

1) a functional relationship between input and

output of the system – map
2) a state-space description of the system
– ordinary differential equation
 1) Input-Output Representation
The mathematical representation of a system can
be described by a functional relationship between
input and output:

Input
u(t) G Output
y(t)

Relationship between y and u is described by an

operator (map) G:
y = G[u]
Example: y = G[u] = 3u
 Transfer Function

Laplace Transforms:

Input
U(s) G Output
Y(s)

Transfer F unction :
Laplace Transform of Output Y ( s)
G(s) =
Laplace Transform of Input U ( s)

(with zero initial conditions)

Example:

a
y + by + cy = du
Laplace
transforms:
Y L{ y}, U L{u}

as 2Y bsY cY dU

(as 2 bs c)Y dU

d
Y 2
U i.e., Y = GU
as bs c
Laplace Transform converts differential equation
description to transfer function description:
a
y + by + cy = du

u(t) y(t)
Plant

U(s) d Y(s)
2
as + bs + c
Computing Output Reponse

The diagram

Fourier Analysis 4-7

Exercise: (5 minutes)

Given:
ay by cu du (with zero initial conditions)

Find: Transfer Function

Y ( s)
G( s) ?
U ( s)
Answer:
𝑎𝑠 ! 𝑌 𝑠 − 𝑏𝑌 𝑠 = 𝑐𝑠𝑈 𝑠 + 𝑑𝑈 𝑠
à
!(#) &#'(
𝐺 𝑠 = %(#) = )#! *+
 Example: Mass-Spring-Damper System
y

u
m
y + cy + ky = u
input force

Taking Laplace Transform (with zero initial conditions)

(ms2+cs+k)Y(s) = U(s)
yields the transfer function 1
G(s) = 2
ms + cs + k
Input-output relationship of the system:

U(s) Y(s)
G(s)

Y(s) = G(s)U(s)
Hence,
1
y (t ) L {G ( s )U ( s )}

(Output obtained by Inverse Laplace Transform)

 Reduction Rules for Complicated Systems

(1) Cascaded system

C/R = G1G2
Summing Point

(2) Parallel system

C/R = G1+G2

(3) Feedback (closed-loop) system

C(s) G1 (s)
=
R(s) 1+ G1 (s)G2 (s)
Block Diagram Operation

Fourier Analysis 4-12

Block Diagram Operation

Fourier Analysis 4-13

 Verifying
E(s)

𝐶 𝑠 = 𝐺! 𝑠 𝐸 𝑠
𝐸 𝑠 = 𝑅 𝑠 − 𝐺" 𝑠 𝐶(𝑠)

𝐶(𝑠)
= 𝐸 𝑠 = 𝑅 𝑠 − 𝐺" 𝑠 𝐶 𝑠
𝐺! 𝑠

𝐶 𝑠 + 𝐺! 𝑠 𝐺" 𝑠 𝐶 𝑠 = 𝐺! 𝑠 𝑅 𝑠
𝐶(𝑠) 𝐺! (𝑠)
=
𝑅 𝑠 1 + 𝐺! 𝑠 𝐺" (𝑠)
Exercise:

C ( s)
? Answer:
R( s )
 Linear Time-Invariant (LTI) Systems
Let a system be described by an operator G as follows:
y(t) = G[u(t)]
Definition of a Linear System
The system described by the operator G is linear, if given
y1(t) = G[u1(t)]
y2(t) = G[u2(t)]
then for any constants a and b, one has
G[a u1(t)+b u2(t)] = a y1(t) + b y2(t)

Definition of a Time-Invariant System

System input-output relationship is time-independent
or
System can be represented by differential equations with
constant coefficients
Case Study 1:
Input(s)
u(t) G Output(s)
y(t)

𝑑𝑦
𝐺: = 2𝑢 𝑡 ⇒ 𝑦 𝑡 = 2 2𝑢(𝑡)𝑑𝑡
𝑑𝑡
Let y(t) = f(u(t)). Then

𝑓 𝛼𝑢! 𝑡 + 𝛽𝑢" 𝑡 = 2(𝛼2𝑢!(𝑡) + 𝛽2𝑢" 𝑡 )𝑑𝑡

= 2 𝛼2𝑢! 𝑡 𝑑𝑡 + 2 𝛽2𝑢" 𝑡 𝑑𝑡 = 𝛼 2 2𝑢! 𝑡 𝑑𝑡 + 𝛽 2 2𝑢" 𝑡 𝑑𝑡

= 𝛼𝑓 𝑢! 𝑡 + 𝛽𝑓 𝑢" 𝑡
è It is linear
All coefficients are constants (for example, that of 𝑢(𝑡) is 2)
è System is time-invariant
Case Study 2:
Input(s)
u(t) G Output(s)
y(t)

𝑑𝑦 " "
𝐺: =𝑡 𝑢 𝑡 ⇒ 𝑦 𝑡 = 1𝑡 𝑢 𝑡 𝑑𝑡
𝑑𝑡
Let y(t) = f(u(t)). Then
"
𝑓 𝛼𝑢! 𝑡 + 𝛽𝑢" 𝑡 = ) 𝑡 𝛼𝑢! 𝑡 + 𝛽𝑢" 𝑡 𝑑𝑡

= ) 𝑡 𝛼 " 𝑢!" + 2𝑢! 𝑢" + 𝛽" 𝑢"" 𝑑𝑡

= 𝛼 " 𝑓 𝑢! 𝑡 + 2𝛼𝛽 ) 𝑡𝑢! 𝑡 𝑢" 𝑡 𝑑𝑡 + 𝛽" 𝑓 𝑢" 𝑡 ≠ 𝛼𝑓 𝑢! 𝑡 + 𝛽𝑓 𝑢" 𝑡

è It is nonlinear
"
Also, the term ‘t’ in 𝑡 𝑢 𝑡 is not constant
è System is time-varying (or, time-variant)
Both “nonlinear” and “time-varying” systems will not be studied in this course
Models of Mechanical Structures: Pendulum
Mass: m
Input: force u(t)
Output: horizontal displacement h(t) (measurement)

Direction of motion
𝜏
𝑇 = 𝑚𝑔 sin 𝜃

h(t) 𝜏 = 𝑚𝐿𝜃̈
𝜃
T 𝜏 = 𝜏! + 𝜏"
𝜏! = −𝑇 = −𝑚𝑔 sin 𝜃
𝑚𝑔 𝜏" = 𝑢 𝑡 cos𝜃

𝑚𝐿𝜃̈ = −𝑚𝑔 sin𝜃 + 𝑢 cos𝜃

𝑚𝐿𝜃̈ = −𝑚𝑔 sin𝜃 + 𝑢 cos𝜃
h(t)

Laplace Transform 𝕃{ A }

𝑚𝐿𝑠 "Θ 𝑠 = −𝑚𝑔 𝕃{sin 𝜃 𝑡 ) + 𝕃{𝑢 𝑡 𝑐𝑜𝑠(𝜃 𝑡 )}

This involves “nonlinear” functions

This types of “nonlinear” systems will not be studied in this course
 Linear Mechanical System: Mass-Spring-Damper

u
m
y + cy + ky = u
input force

Taking Laplace Transform (with zero initial conditions)

(ms2+cs+k)Y(s) = U(s)
yields the transfer function 1
G(s) = 2
ms + cs + k
 Single-Input Single-Output (SISO) Systems
A state representation of a given transfer function is not
unique, namely, it may have different forms.
Given a system transfer function:
Y (s) b0 s m b1s m 1 bm 1s bm
n n 1
U (s) s a1s an 1 s an
For physical systems, because of casualty and delay, m < n
Some models have same order of input and output, m = n
SISO refers to the dimension, not the order (m or n),
of the input-output (i.e., both input and output are
scalers but not vectors)
 Models of Electric Circuits

Basic Equations of Electric Circuits:

• Kirchhoff’s Current Law

– The algebraic sum of currents leaving a junction
(a node) is equal to the algebraic sum of currents
entering that node
• Kirchhoff’s Voltage Law
– The algebraic sum of all voltages taken around a
closed path in a circuit is 0
 Kirchhoff’s Current Law

Kirchhoff’s Voltage Law

 Impedance
Resistor Inductor Capacitor
(with zero initial current) (with zero initial voltage)

v(t ) Ri (t ) di (t ) dv (t )
v(t ) L i (t ) C
dt dt
V ( s) RI ( s )
V (s) LsI ( s ) I ( s) CsV ( s )
V ( s)
R V (s) V (s) 1
I (s) Ls
I ( s) I ( s) Cs

Impedance

Fourier Analysis 4-25

 RC Circuit

𝑒! − 𝑒"
𝑖=
𝑅

∫ 𝑖𝑑𝑡
𝑒" =
𝐶

Parameters: R, C

Independent variables (inputs): 𝑒'

Dependent variables (outputs): 𝑒( and 𝑖
Laplace Transforms:
 LRC Circuit

Transfer Function:

#! $" #$"
[ Time Domain: 𝐿𝐶 #% ! + 𝑅𝐶 #%
+ 𝐸& = 𝐸' ]
 Transfer Function of a Cascaded Network

Example:

Steps: (1) find the circuit equations

(2) find their Laplace Transforms
(3) find the overall transfer function
(4) build the block diagram
First, find the circuit equations
𝑉! − 𝑉#$ 𝑉#$ − 𝑉"
𝑖= 𝑖% = 𝑖 = 𝑖! + 𝑖"
𝑅$ 𝑅%

∫ 𝑖$ 𝑑𝑡 ∫ 𝑖% 𝑑𝑡
𝑉#$ = 𝑉" =
𝐶$ 𝐶%
 𝑉! − 𝑉#$ 𝑉#$ − 𝑉"
𝑖= 𝑖% = 𝑖 = 𝑖! + 𝑖"
𝑅$ 𝑅%
∫ 𝑖$ 𝑑𝑡 ∫ 𝑖% 𝑑𝑡
𝑉#$ = 𝑉" =
𝐶$ 𝐶%

Then, find their Laplace Transforms

𝑉& 𝑠 − 𝑉#$ (𝑠) 𝑉#$ 𝑠 − 𝑉" (𝑠)

𝐼 𝑠 = 𝐼% 𝑠 = 𝐼 𝑠 = 𝐼$ 𝑠 + 𝐼% 𝑠
𝑅$ 𝑅%

𝐼$ (𝑠) 𝐼% (𝑠)
𝑉#$ 𝑠 = 𝑉" 𝑠 =
𝐶$ 𝑠 𝐶% 𝑠
 𝑉 𝑠 − 𝑉#$ (𝑠) 𝑉#$ 𝑠 − 𝑉" (𝑠)
𝐼 𝑠 = & 𝐼% 𝑠 = 𝐼 𝑠 = 𝐼$ 𝑠 + 𝐼% 𝑠
𝑅$ 𝑅%

𝐼$ (𝑠) 𝐼% (𝑠)
𝑉#$ 𝑠 = 𝑉" 𝑠 =
𝐶$ 𝑠 𝐶% 𝑠

Then, find the overall transfer function

𝑉#$ 𝑠 − 𝑉" (𝑠)

𝐼 𝑠 = 𝐼$ 𝑠 + 𝐼% 𝑠 𝐼% 𝑠 =
𝑅%

𝑉& 𝑠 − 𝑉#$ (𝑠) 𝑉#$ 𝑠 = 𝑅% 𝐼% 𝑠 + 𝑉" 𝑠

= 𝐶$ 𝑠𝑉#$ 𝑠 + 𝐶% 𝑠𝑉" 𝑠
𝑅$ = 𝑅% 𝐶% 𝑠 𝑉" 𝑠 + 𝑉" 𝑠

𝑉( 𝑠 = 𝑅! 𝐶! 𝑅" 𝐶" 𝑠 " + (𝑅! 𝐶! + 𝑅" 𝐶" + 𝑅! 𝐶" 𝑠 + 1]𝑉& (𝑠A

[ #! )"
Time domain: 𝑅! 𝐶! 𝑅" 𝐶" #% ! + 𝑅! 𝐶! + 𝑅" 𝐶" + 𝑅! 𝐶"
#)"
#%
+ 𝑉& = 𝑉( ]
 Finally, build the block diagram

𝑉( 𝑠 = 𝑅! 𝐶! 𝑅" 𝐶" 𝑠 " + (𝑅! 𝐶! + 𝑅" 𝐶" + 𝑅! 𝐶" 𝑠 + 1]𝑉& (𝑠A

𝑉" (𝑠) 1
=
𝑉& (𝑠) 𝑅$ 𝐶$ 𝑅% 𝐶% 𝑠 % + (𝑅$ 𝐶$ + 𝑅% 𝐶% + 𝑅$ 𝐶% 𝑠 + 19

A second-order system
 Framework of Lectures
System

Time Domain Frequency Domain

à Laplace Transform à

State-Space Formulation Frequency Analysis

This course is mainly about

engineering systems described by
Second-Order Ordinary Differential Equations
 For a system with input and output

System

Typically it can be described by a second-order ODE

Example:

Time domain:

𝑑 ! 𝐸" 𝑑𝐸"
𝐿𝐶 ! + 𝑅𝐶 + 𝐸" = 𝐸#
𝑑𝑡 𝑑𝑡

Frequency domain:
 For a system with input and output

System

Typically it can be described by a second-order ODE

Example:
y Time domain:

m
y + cy + ky = u
u
input force Frequency domain:
1
G(s) = 2
ms + cs + k
 DC Motor:
Electro-Mechanical System

𝑉 𝑡 = Armature Voltage 𝜃 𝑡 = Angular Position

𝑖 𝑡 = Armature Current 𝐽 = Rotor Inertia
𝑒 𝑡 = Back EMF 𝐵 = Viscous Friction
𝐾 = Motor Constant
 𝜏

The DC motor armature circuit is described based on the Kirchhoff voltage law by

#'(%)
𝑉 𝑡 = 𝑅𝑖 𝑡 + 𝐿 + ̇
𝐾 𝜃(𝑡)
#%

The mechanical load is described via the torque balance by

̈
𝐽𝜃(𝑡) ̇
= 𝜏(𝑡) − 𝐵𝜃(𝑡)

where
𝑉 𝑡 = Armature Voltage 𝜃 𝑡 = Angular Position
𝑖 𝑡 = Armature Current 𝐽 = Rotor Inertia
𝑒 𝑡 = Back EMF 𝐵 = Viscous Friction
𝐾 = Motor Constant
Motor torque is proportional to the armature current:
𝜏(𝑡) = 𝐾' 𝑖(𝑡)

Back EMF is proportional to the angular velocity of the shaft:

̇
𝑒(𝑡) = 𝐾 𝜃(𝑡)
Assuming that the motor torque and the back EMF constants
are equal: 𝐾, = 𝐾, the above governing equations become
𝜏(𝑡) = 𝐾𝑖(𝑡)
𝑇 𝑠 =?
̇
𝑒(𝑡) = 𝐾 𝜃(𝑡) Laplace
Transforms 𝐸 𝑠 =?
#'(%)
Exercise
𝑉 𝑡 = 𝑅𝑖 𝑡 + 𝐿 #% + ̇
𝐾 𝜃(𝑡) 𝑉 𝑠 =? (5 Minutes)
𝐽𝜃̈ 𝑡 = 𝐾𝑖 𝑡 − 𝐵𝜃(𝑡)
̇ Θ(𝑠) = ?

( )
H s =* =?
+
Motor torque is proportional to the armature current:
𝜏(𝑡) = 𝐾' 𝑖(𝑡)

Back EMF is proportional to the angular velocity of the shaft:

̇
𝑒(𝑡) = 𝐾 𝜃(𝑡)
Assuming that the motor torque and the back EMF constants
are equal: 𝐾, = 𝐾, the above governing equations become
𝜏(𝑡) = 𝐾𝑖(𝑡) 𝑇 𝑠 = 𝐾𝐼 𝑠
̇
𝑒(𝑡) = 𝐾 𝜃(𝑡) Laplace
Transforms 𝐸 𝑠 = 𝐾𝑠Θ 𝑠
#'(%)
𝑉 𝑡 = 𝑅𝑖 𝑡 + 𝐿 + ̇
𝐾 𝜃(𝑡)
#% 𝑉 𝑠 = 𝑅 + 𝐿𝑠 𝐼 𝑠 + 𝐾𝑠Θ (𝑠)

𝐽𝜃̈ 𝑡 = 𝐾𝑖 𝑡 − 𝐵𝜃(𝑡)
̇ 𝑠(𝐽𝑠 + 𝐵)Θ(𝑠) = 𝐾𝐼 𝑠

Θ s 𝐾
H s = =
𝑉 𝑠 𝑠 𝑅 + 𝐿𝑠 𝐽𝑠 + 𝐵 + 𝐾 %
 𝑉 𝑠 − 𝐾𝑠Θ(𝑠)
𝐼 𝑠 =
𝑅 + 𝐿𝑠
𝑉 𝑠 − 𝐾𝑠Θ(𝑠)
𝑠(𝐽𝑠 + 𝐵)Θ(𝑠) = 𝐾
𝑅 + 𝐿𝑠

Θ s 𝐾
H s = =
𝑉 𝑠 𝑠 𝑅 + 𝐿𝑠 𝐽𝑠 + 𝐵 + 𝐾 %

Block diagram of the DC motor from input 𝑉(𝑠) to output Θ(𝑠):

 Framework of Lectures
System

Time Domain Frequency Domain

à Laplace Transform à

State-Space Formulation Frequency Analysis

(dynamic responses, transients,
steady-states, root-locus,
PID controllers)
ß mixed à
(State-feedback,
Controllability, controller This course is mainly about
Observability, observer engineering systems described by
Optimal control)
Second-Order Ordinary Differential Equations
Classification of Systems
q Memoryless

q Causality

q Stability

q Time-invariance

q Linearity

Systems 2-43
(1). Memoryless

q A system is memoryless if the output at ANY

given time depends ONLY on the input value
at that (same) time.

q A system with memory is called a dynamic

system.

Systems 2-44
Examples
1. A resister is a memoryless component:
(1) y(t) = R x(t)
(2) y(t) = x(t-1), y(t) = x(t+1), y(t) = x(2t)
(3) y(t) = sin[x(t)], y(t) = [x(t)]^2
(4) y(t) = t x( )d
2. Inductors and capacitors are components having
infinite memory.

3. The following system has finite memory:

y[n] x[n] x[n 1] x[n 2]
Systems 2-45
(3). Causality
q A system is causal if the output at ANY time depends
ONLY on the input values at and prior to that time.

Past, Present, and Future

q Examples
m Causal:
y[n] x[n] x[n 1]
m Non-causal:
1
y[n] x[n 1] x[n] x[n 1]
3
Systems 2-46
Examples of Causal & Non-Causal Systems

(1) y(t) = R x(t)

(2) y(t) = x(t-1), y(t) = x(t+1), y(t) = x(2t)
(3) y(t) = sin[x(t)], y(t) = [x(t)]^2
(4) y(t) = t x( )d
dx (t )
(5) y (t )
dt

qObservation: Every memoryless system is causal.

Systems 2-47
Remarks on Causality

q Causality is a must for real-time processing

and implementation, which governs the
fundamental principle of cause-and-effect.
q Any system that is not causal cannot be
realized by physical devices; a system that
is causal can in principle be built.
Examples: A differentiator cannot be built
in reality, an integrator can be readily built
using, e.g., an Op-Amp.

Systems 2-48
(4). Stability
q Informally, a system is stable if small inputs
lead to responses (i.e. outputs) that do not
diverge.

x(t) : the horizontal force; y(t) : vertical displacement

x(t) y(t)

y(t)

x(t)

Stable System Unstable System

Systems 2-49
(4). Stability (BIBO stable)
q Formally, a system is stable if every bounded
input results in a bounded output.

q Given that the input signal has an upper bound

A, i.e.,
x(t ) A for all t ,

there exists a number B such that

y(t ) B for all t.

Systems 2-50
Examples

1. Is the squaring system y(t) = x2(t) stable?

Yes

n
2. Is the accumulator system y[n] x[k ] stable?
k 1

Hint: Consider the input x[n] = {1, 1, 1, 1, …}.

No

Systems 2-51
Examples of Stable Systems

(1) y(t) = R x(t) stable

(2) y(t) = x(t-1), y(t) = x(t+1), y(t) = x(2t) stable
(3) y(t) = sin[x(t)], y(t) = [x(t)]^2 stable
(4) y(t) = t x( )d unstable
dx (t )
(5) y (t ) unstable
dt

qProving stability: Show for all bounded inputs the output is

bounded.
qDisapproving stability: Construct one bounded input and show
that the corresponding output is unbounded-one strike rule.
Systems 2-52
(5). Time Invariance

q A system is time invariant if a time shift in

the input causes the SAME time shift in
the output:
If x[n] ® y[n], then x[n-n0] ® y[n-n0];
If x(t) ® y(t), then x(t-t0) ® y(t-t0).
The system doesn’t change with time.
q A system that is not time invariant is said
to be time varying.

Systems 2-53
Example

q Is the system described by y[n] = n x[n]

time invariant?

No

Systems 2-54
Examples of Time-Invariant Systems

(1) y(t) = R x(t)

(2) y(t) = x(t-1), y(t) = x(t+1), y(t) = x(2t)
(3) y(t) = sin[x(t)], y(t) = [x(t)]^2
(4) y(t) = t x( )d

qProving and disapproving time invariance: Construct a shifted

input x(t-t0), use the system definition, see if you can get y(t-t0).

Systems 2-55
(6). Linearity

q Let x1(t) ® y1(t) , x2(t) ® y2(t) .

q A system is linear if it satisfies these conditions:

1. x1(t)+x2(t) ® y1(t) +y2(t) --- additivity

2. a x1(t) ® a y1(t), where a is any complex constant.

--- scaling (homogeneity)

Systems 2-56
Superposition Property

q Taking the two properties together, we have

a x1(t)+ b x2(t) Þ a y1(t) + b y2(t)

--- superposition property

q A system is linear if it satisfies the

superposition property.

Systems 2-57
Example

1. Consider a system described by

y(t) = t x(t).
Is it linear?
Yes

q Hint: Procedure to test linearity:

m Let y1(t) = t x1(t) and y2(t) = t x2(t).
m Consider the input x(t) = a x1(t) + b x2(t).
m Find the corresponding output y(t).
m The system is linear iff y(t) = a y1(t) + b y2(t).

Systems 2-58
Example

2. Consider a system described by

y[n] = 2 x[n] + 3.
Is it linear?

No, because it does not have the zero-

in zero-out property.

Note: This system belongs to a class called affine or

incrementally linear system.

Systems 2-59
END