MA3005: Control Theory

School of Mechanical and Aerospace

Engineering

Faculty: Asst Prof Tegoeh Tjahjowidodo

Email: [email protected]
Phone: (+65)6790 4952
Office: N3-02c-68
Week 4 –Response and Stability
Part I – Lecture Map
2. Laplace
Transform
3. Modelling

1. Introduction to Systems

6. First Order 4. Response and

Stability

5. Routh Stability
8. Transient 7. Second Order
Response

9. Basic of PART II. CONTROLLER DESIGN

Control

Week 4 3
Learning Objectives
At the end of this week, you should be able to:
• Identify poles and zeroes
• Assess system stability

Transient Steady-state
response response

y(t)
1

x(t)

0 10 20 30 Time (s)

Week 4 4
Time Response
Given a system:

R(s) C(s)
G(s)

• How to assess the performance of the system?

• How far should we design our controller?

• How good is our system with a controller?

Week 4 5
Recall – Dynamic Systems
Dynamic systems are characterised mathematically by derivative and/ or integrative operators.
Example: 5 dx  x  y (5s  1) X  Y
X

1
dt Y 5s  1
d 2x dx X 1
 5  10 x  y (s 2  5s  10) X  Y  2
dt 2
dt Y s  5s  10

𝒀
y(t) x(t)
Input 1

Rule
1
5s  1
Output
0 10 20 30 time (s)
Time (s)

𝑿
Tool for evaluating response: Laplace Transform
Week 4 6
Response of Dynamic Systems
How does the system behave for any arbitrary input, x(t)?

𝒀
y(t)
Input 1

Rule
1
x(t)???
5s  1
Output
0 10 20 30 Time (s)
𝑿

Week 4 7
Linear Dynamic System
In this course, we are only focusing on linear dynamic system.
A linear dynamic system holds the superposition principle, i.e., the response of the system
caused by two or more inputs is equal to the sum of the responses (of each input).

u1(t) y1(t)
G(s)
u1(t)+u2(t) y1(t)+y2(t)
u2(t)
 G(s)
y2(t)
G(s)

Week 4 8
Application of Superposition
A complex input signal can be broken down into basic signals.
Response can be evaluated individually and later summed to get response to original input.

t1 t2 t3 t4

Week 4 9
Test (Input) Signals
Superposition brings advantage: system characterisation can be carried out sufficiently by
analysing particular responses.
Typical test signals: step, ramp, parabolic, impulse, and sinusoidal function.
Step Ramp Parabolic
u(t) u(t) u(t)

Time, t Time, t Time, t

Impulse Sinusoidal
u(t)
u(t)

Time, t
Time, t

Week 4 10
How to Analyse Responses?
System Response
System response to any test signal can be evaluated easily in Laplace domain:

u(t) y(t)
G(s)

Y ( s )  G ( s )  U ( s ) where: Y ( s )  L [ y (t )]
U ( s )  L [u (t )]

Week 4 12
System Response (Cont’d.)
For a linear system:

Total response = Natural response + Forced response

(Recall MA2006 on differential equation: homogeneous and particular solutions)

Week 4 13
Time Response Plot (Cont’d.)
Consists of two parts: transient and steady-state responses

Transient Steady-state
response response

y(t)
1

x(t)

0 10 20 30 Time (s)

Week 4 14
Time Response Plot (Cont’d.)
Consists of two parts: transient and steady-state responses

Transient Steady-state
response response

y(t) x(t)
1

0 10 20 30 Time (s)

Week 4 15
Time Response Plot (Cont’d.)
Transient response is dominated by natural response.
Steady-state response is dominated by forced response.
Time response can be considered combination of the forced and natural responses or
combination of transient and steady-state responses.

c(t) = cforced(t) + cnatural(t) = ct(t) + css()

Week 4 16
Dynamic Response
System accuracy is determined from steady-state response. Transient Steady-state
response response
System speed is inferred from transient response. y(t)
1
• A faster system is able to track input commands better x(t)
(tracking ability).
0 10 20 30 Time (s)
This week mainly discusses the steady-state response.
Transient response will be discussed in topics on First Order,
Second Order, and Transient Response.

Week 4 17
Steady-State Response
Steady-state response, in some cases, can be evaluated using Transient Steady-state
response response
the Final Value Theorem:
y(t)
u(t) y(t) 1
x(t)
G(s)
0 10 20 30 Time (s)
lim y (t )  lim s  Y ( s )
t  s 0
 lim s  G ( s )  U ( s ) where: Y ( s )  L [ y (t )]
s 0
U ( s )  L [u (t )]

Week 4 18
DC Gain (Zero Frequency Gain)
By definition, DC gain is the steady-state response to a unit step input. u(t)

From the Final Value Theorem: 1

DC gain  lim y (t )  lim s  Y ( s ), where Y(s) is the step response:

t  s 0 Time, t

1
Y (s)  U (s)  G(s) Y (s)   G (s)
s
= U(s)
Therefore, 1
DC gain  lim s  G ( s )
s 0 s

DC gain  lim G ( s )  G ( s ) s 0
s 0

Week 4 19
Example
Given the following system:

3( s  2)
G (s)  2
s  2 s  10
3 2
DC gain  G ( s ) s 0   0.6
10
i.e. for step input of 5u(t), the steady-state is:

css() = 5·0.6 = 3

Week 4 20
Closed-Loop System Error

U(s) E(s) C(s)

G1(s)
+
-

An error is the difference between the reference signal and output signal.
An ideal control system should give C = U. That is the reason to refer the difference as an error.
When we are discussing about a closed-loop system, it is necessary to quantify the controlled
system performance, in terms of the steady-state of the error signal.

Week 4 21
System Error
System error applies to a controlled system.
Let’s consider a unity feedback system: U(s) E(s) C(s)
G1(s)
E (s)  U (s)  C (s) +
-
From the feedforward part,
C ( s )  G1 ( s )  E ( s )
Therefore,
E ( s )  U ( s )  G1 ( s )  E ( s )

1
E (s)   U (s)
1  G1 ( s )

Week 4 22
System Error (Cont’d.)
System error applies to a controlled system.
Let’s consider a unity feedback system: U(s) E(s) C(s)
G1(s)
Alternatively, the closed-loop system can be +
represented as a closed-loop transfer -
function (CLTF):
G1 ( s )
G(s)  =
1  G1 ( s )
U(s) G1 ( s ) C(s)
As E ( s )  U ( s )  C ( s ) where C ( s )  G ( s )  U ( s ), 1  G1 ( s )

therefore, = G(s)
E ( s )  (1  G ( s ))  U ( s )

Week 4 23
Steady-State Error
Steady-state error represents the final value of the error.
From the Final Value Theorem:
ess (t )  lim e(t )  lim s  E ( s )
t  s 0
1
 lim s   U (s)
s 0 1  G1 ( s )
or  lim s  (1  G ( s ))  U ( s )
s 0

Week 4 24
Steady-State Error (Cont’d.)
In case of a system with unit step input, the steady-state error is:
From the Final Value Theorem:

ess (t )  lim e(t )  lim s  E ( s )

t  s 0 = R(s)
1 1 1
 lim s    lim
s 0 1  G1 ( s ) s s 0 1  G1 ( s )
1
or  lim s  (1  G ( s ))   lim(1  G ( s ))
s 0 s s 0

Week 4 25
Example
The system G1(s) is used in a unity feedback system: U(s) E(s) C(s)
G1(s)
3( s  2) +
G1 ( s )  2 -
s  2s  10
The error to a step input of 5u(t) is:
1 s 2  2s  10 50
ess  5   5 2 
1  G1 ( s ) s 0
( s  2s  10)  3( s  2) s 0
16

Week 4 26
How does the System Property (Poles) Affect the Response?
Recall: Week 2 – Illustration of System Model (1)
Consider a dynamic system defined by the differential equation:
(n) ( n 1) (m) ( m 1)
a0 y  a1 y    an 1 y  an y  b0 x  b1 x    bm 1 x  bm x
where y is the output and x is the input.

In Laplace domain (with zero initial condition), it can be written as:

a0 s nY  a1s n 1Y    an 1sY  anY  b0 s m X  b1s m 1 X    bm 1sX  bm X

Y ( s ) b0 s m  b1s m 1    bm 1s  bm
or  n 1
 G(s)
X ( s ) a0 s  a1s    an 1s  an
n

where G(s) is the transfer function of the system.

Week 4 28
Recall: Week 2 – More on Transfer Function
b0 s m  b1s m 1    bm 1s  bm
A transfer function G ( s )  n 1 can be written in factorised pole-zero
a0 s  a1s    an 1s  an
n

form:
K  s  z1  s  z2   s  zm   N  s 
G( s) 
 s  p1   s  p2   s  pn  D  s 
• Poles: –p1, –p2, … are values of s that make the transfer function G(s) singular or |G(s)|
infinite (= roots of denominator).
• Zeros: –z1, –z2, … are values of s that make the transfer function |G(s)| = 0 (= roots of
numerator).
• When n > m, then G(s) has 𝑚 finite zeros and (n – m) zeros at s = ∞ (infinite zeros).

Week 4 29
Pole-Zero Form of Transfer Function
Consider a transfer function:
K  s  z1  s  z2   s  zm   N  s  N(s) = numerator polynomial
G( s) 
 s  p1   s  p2   s  pn  D  s  D(s) = denominator polynomial

Some of the poles (as well as the zeros) might appear in complex forms. However, the
conjugate of the complex pole (or zero) will also appear as another pole (or zero) on the
transfer function.
A transfer function with complex poles:
K ( s  z1 )( s  z 2 )  ( s  z m )
G (s)  r
 ( s   )  n ( s   )( s   ) 
 i   i 
 i 1  i  r 1 
i

where i and i are complex conjugates.

Week 4 30
Example
What are the poles of the following system?
K ( s 2  4s  3) K ( s  1)( s  3) K ( s  1)( s  3)
G ( s)  4  
s  8s  45s  116s s ( s  4)( s  4 s  29) s ( s  4)( s  2  5 j )( s  2  5 j )
3 2 2

• Two first order poles at s = 0, –4

• A pair of complex conjugate poles at –2 + 5j (second-order poles)

Week 4 31
Defining Poles (Alternative)
Consider a transfer function:
N ( s) numerator polynomial
G( s)  
D( s) denominator polynomial

• The characteristic equation is the denominator polynomial set to zero: D(s) = 0

• The roots of the characteristic equation are poles

Week 4 32
Poles
Poles determine the stability of a system and the general nature of a transient response to any
input.
Goes to infinity
Example: Effect of a Pole
Consider the unit step response of:
C s
a>0
a
G( s)  
R s s  a
0 Time (s)
For the unit step response:
a 1 1 1
C s   
sa s s sa a<0

 c  t   1  e at 0 Time (s)
Response increases to ∞ for a > 0.
Response settles to –1 for a < 0. Converges to a final value
Poles Location Map in a Complex Plane

Week 4 33
Natural Response
(a) Stable Converging

A system is called stable if the output of the system

converges to a particular value/ state for a constant input. t

Goes to infinity
(b) Unstable
When a system is unstable, the output of the system
goes to infinity for finite input (or constant input). t

(c) Marginally Stable Oscillating

A marginally stable system will have continuous oscillatory

output for a constant input. t

Week 4 34
Remarks
Poles can be real or complex, as well as zeros.
If complex, they (always) occur in complex conjugate pairs.
Dynamic response (and stability) of a system depends on the location of the poles of the
system.

Week 4 35
What is the Effect of Pole Position to the Response?
Complex Plane (z-plane/ s-plane)
The complex plane or z-plane is a geometric representation of the complex numbers
established by the real axis and the orthogonal imaginary axis.

Imaginary
j
axis

3
–3 + 2j
2
2+j
1
3
Real axis
–4 –3 –2 –1 0 1 2 3 4
–1

–2

–3

Week 4 37
Poles Location Map in a Complex Plane
Imaginary
jω
axis

When the complex plane is used to

1 1 1
present poles/ zeros locations, it is
usually referred to as an s-plane.

Real axis
σ
1 1 1
0 0
t
LHP = Left Hand Plane RHP = Right Hand Plane Return to Poles
Week 4 38
Example
1
G(s)  First-order pole at s = – 4
s4
The step response of the system, G(s):
1 1 1 11 1 
C (s)  G(s)      
s s4 s 4 s s4

c(t )  L -1[C ( s )] 
1
4

1  e  4t  1
x(t)
4

0 10 20 30 Time (s)

Week 4 39
Poles Location and Natural Responses
Imaginary jω
1
G(s)  axis
s4
The system is stable.
There is no oscillation on the response.
1 1 1

Real axis

1 1 1 σ
0 0
t

Week 4 40
Example
1
G(s)  2 Second-order poles at s = –2 + 5j and s = –2 – 5j
s  4s  29
The step response of the system, G(s):
1 1 1 1 29 1 2 2  52
C (s)  G (s)  2     
s s  4s  29 s 29 s ( s  4s  29) 29 s[( s  2) 2  52 ]
2

c(t )  L [C ( s )] 
-1 1
29
 
1  e  2t (cos 5t  52 sin 5t )

x(t)
1
29

0 0.2 0.4 0.6 Time (s)

Week 4 41
Poles Location and Natural Responses
Imaginary jω
1
G(s)  2 axis
s  4 s  29
The system is stable.
Oscillation is observed on the response.
1 1 1

Real axis

1 1 1 σ
0 0
t

Week 4 42
Example
1
G (s)  2 Second-order poles at s = 2j and s = –2j
s 4
The step response of the system, 𝐺(𝑠):
1 1 1 1 1 s 
C (s)  G (s)  2     2
s s  4 s 4  s s  2 2 

c(t )  L [C ( s )]  1  cos 2t 
-1 1
4
1 x(t)
4

0 3 6 9 12 Time (s)

Week 4 43
Poles Location and Natural Responses
Imaginary jω
1
G (s)  2 axis
s 4
The system is marginally stable.
The response is not converging nor going to infinity.
1 1 1

Real axis

1 1 1 σ
0 0
t

Week 4 44
Routh-Hurwitz – Part 1
Part I – Lecture Map
2. Laplace
Transform
3. Modelling

1. Introduction to Systems

6. First Order 4. Response and

Stability

5. Routh Stability
8. Transient 7. Second Order
Response

PART II. CONTROLLER DESIGN

Week 4 46
Poles Locations and Natural Responses
𝑗𝜔

1 1 1

𝜎
1 1 1
0 0
𝑡

Week 4 47
Stability Analysis in a Complex Plane
Stability of linear closed-loop system can be determined from the location of the closed-loop
system poles in s-plane.
Unstable systems have at least one pole in the RHP and transient response that increases
monotonically or oscillates with increasing amplitude.

b0 s m  b1s m 1    bm 1s  bm
G ( s) 
a0 s n  a1s n 1    an 1s  an

Week 4 48
Stability Conditions
jw The conditions are:
• Stable systems have all closed-loop poles lie to the left of
the jw-axis (necessary and sufficient condition).
s-plane
• Poles of the CLTF have negative real parts.

Week 4 49
Recall: Final Value Theorem
lim f (t )  lim sF ( s )
t  s 0
Properties of the Final Value Theorem:
• Relates steady-state behaviour of f (t) to behaviour of sF(s) in neighborhood of s = 0
• Applies only if lim f (t ) exists
t 

• That is, all poles of F(s) lie in the left half s-plane (and at most one pole on the imaginary
axis)

Week 4 50
Effect of Input on System
Stability is property of system and is independent of input or driving function.
Poles of the input do not affect stability of system but contribute to steady-state response.

Week 4 51
Determining Stability
N ( s) numerator polynomial
D(s) = 0 G( s)  
D( s) denominator polynomial

We solve the characteristic equation and locate the poles in the complex s-plane.
However, sometimes, it is not straightforward to find the roots of higher order polynomials.

Examples:
Defining stability from the characteristic equations:
• s2 + 4s + 3 = 0 (s + 1)(s + 3) = 0
s1 = –1 and s2 = –3 (all negative, therefore stable!)
• s2 + 2s – 3 = 0 (s – 1)(s + 3) = 0
s1 = +1 and s2 = –3 (one positive, therefore unstable!)

Week 4 52
Determining Stability (Cont’d.)
s6 + 7s5 + 18s4 + 9s3 + 8s2 + 4s + 3 = 0 (s + …)(s + …)… = 0
s1 = ? s2 = ? … s7 = ? (stability?)

It is not easy to factorise higher order polynomials.

Methods used are:
• Routh-Hurwitz method: Algebraic method to test whether the roots of the closed-loop
characteristic equation lie in LHP, on jw-axis or in RHP
• Root locus plot: Diagram of loci of characteristic equation roots when a certain parameter
(root locus gain K) varies

Week 4 53
Stability and Characteristic Equation
C ( s ) N ( s ) b0 s m  b1s m1   bm 1s  bm
 
R( s ) D( s ) a0 s n  a1s n 1   an 1s  an

D(s) = 0
A necessary (but not sufficient) condition:
If a system is stable, then all coefficients of its characteristic equation are real, are of the same
sign, and none of its coefficient is zero, that is:
• a0, a1, …, an-1, an > 0 (or < 0)
Otherwise, the system is not stable!
• None of the coefficient is missing

Examples of non-stable systems:

• s2 + 4s – 3 = 0 different signs
• s6 + 7s5 + 9s3 + 8s2 + 3 = 0 zero coefficients

Week 4 54
Stability and Characteristic Equation (Cont’d.)
If all the conditions in the previous slide are met, then no conclusion can be derived:
If all coefficients of its characteristic equation are of the same sign and
none of its coefficient is zero, it does not guarantee the stability.
Routh-Hurwitz criteria must be applied, which gives the necessary and sufficient condition for
stability.
This criteria is applied by forming the Routh’s Array from the characteristic equation.

Week 4 55
Routh’s Array
Characteristic equation: a0 sn + a1 sn–1 + a2 sn–2 + … + an = 0

sn a0 a2 a4 …
From characteristic
sn-1 … equation
a1 a3 a5

sn-2 b1 b2 b3 …

sn-3 c1 c2 c3 …
Need to be computed
… …

s1 a 0

s0 b 0

Week 4 56
Computing Routh’s Array Elements
Example of characteristic equation: a0 s5 + a1 s4 + a2 s3 + a3 s2 + a4 s + a5 = 0

s5 a0 a2 a4 From characteristic
equation
s4 a1 a3 a5

s3 b1 b2 b3

s2 c1 c2 c3
Need to be computed
s1 a 0

s0 b 0

Week 4 57
Computing Routh’s Array Elements (Cont’d.)
where:
a1  a2  a0  a3 a1  a4  a0  a5
b1  , b2 
a1 a1
b1  a3  a1  b2 b1  a5  a1  b3
c1  , c2 
b1 b1

Compute array elements till only one element is left in the first column of the last two rows.

Week 4 58
Computing Routh’s Array Elements (Cont’d.)
Arrange the coefficients as:
sn a0 a2 a4 a6  The third row can be evaluated as:


a1a2  a0 a3 a1a4  a0 a5
sn-1 a1 a3 a5 a7 b1  b2 
a1 a1
sn-2 b1 b2 b3 b4 
The fourth row can be evaluated as:
sn-3 c1 c2 c3 c4 
b1a3  a1b2 b1a5  a1b3
c1  c2 
b1 b1



Week 4 59
Routh-Hurwitz Criterion
sn a0 a2 a4 a6  Refer to the first column:
• If there is no changing of sign, then the closed-loop
sn-1 a1 a3 a5 a7 
system is stable.
sn-2 b1 b2 b3 b4  • The number of times the sign changes in the first
column represents the number of poles in RHP.
sn-3 c1 c2 c3 c4 


Week 4 60
Example 1
s1, 2  1 s3, 4  1  j 3

D(s) = s4 + 4s3 + 9s2 + 10s + 4 = (s + 1)2 (s2 + 2s + 4) = 0

s4 4  9  1  10 26
1 9 4 b1  
4 4
s3 4 10
4  4  1 0
b2  4
s2 b1 = 26/4 b2 = 4 4
s1 c1 = 7.54
c1 
 26 / 4   10  4  4
 7.54
s0 d1 = 4 26 / 4

7.54  4  0   26 / 4 
No sign change in the first column ⇒ system is stable d1  4
7.54

Week 4 61
Routh-Hurwitz – Part 2
Special Case 1: If First Element in Any Row is Zero
If first element in any row is zero but others are non-zero, replace 0 with e > 0 and
proceed as before.
If sign above zero (e ) is the same as below, it indicates that there is a pair of imaginary
roots. ⇒ marginally stable system
If sign above zero (e ) is opposite to that of below it, it indicates that there is a sign
change. ⇒ unstable system
In this case, the system is never stable.

Return to Example 2
Week 4 63
Example 2
Consider the following characteristic equation:
s3 + 2s2 + s + 2 = 0
Form the array:
s3 1 1 Observe that:
s2 2 2 • No sign change in the first column ⇒ no poles in the RHP

s1 0 e • Zero in the first column ⇒ a pair of imaginary roots

2e – (2)(0)
s0 2=
e
s3 + 2s2 + s + 2 = (s + 2) (s2 + 1) = 0

Return to Special Case 1

Week 4 64
Example 3
Consider the following characteristic equation:
s3 – 3s + 2 = 0
Form the array:
s3 1 –3

s2 0 e 2
s1 –3 – 2/e For small positive ε, –3 – 2/ε < 0
s0 2
Different signs above and below ε ⇒ unstable system
s3 – 3s + 2 = (s – 1)2 (s + 2) = 0

Week 4 65
Special Case 2: If the Entire Row of Array is Zero
If the entire row of array is zero, there exist pairs of roots, which are mirror images with
respect to the imaginary axis:
i.e. (–s, +s) or (–jw, +jw) or (–s ± jw, +s ± jw)
Unstable Marginally stable Unstable
jw jw jw

s s s

Solution: Form auxiliary polynomial from the row immediately above all zeros row.
The system is never stable either.
Week 4 66
Example 4
Consider the following characteristic equation:
A(s) = s5 + 2s4 + 24s3 + 48s2 – 25s – 50 = 0
Routh’s array of coefficients is:
s5 1 24 –25
s4 2 48 –50 Form an auxiliary polynomial using this row
s3 0 0 All zero elements
s2 ? ?
s1 ? ?
s0 ?

Week 4 67
Auxiliary Polynomial
Auxiliary polynomial is formed from coefficients of the row immediately above all the zero
rows, i.e., s4. 5
s 1 24 –25
s4 2 48 –50

The auxiliary polynomial P(s) is P(s) = 2s4 + 48s2 – 50, indicating two pairs of roots of equal
magnitude and opposite signs, which can be obtained by solving P(s) = 0.

Week 4 68
Routh’s Array for All Zero Rows
Differentiate P(s) with respect to s:
dP( s )
P(s) = 2s4 + 48s2 - 50  8s 3  96s
ds
Replace all zero rows by dP(s)/ds:
Replace all zero row by dP(s)/ds
s5 1 24 –25 Observe that:
s4 2 48 –50 • The first column of the new array has one sign change ⇒ unstable with
one RHP pole
s3 08 0 96
• The one RHP pole can also be verified from the given characteristic
s2 24 –50 equation:
s1 112.7 0 s5 + 2s4 + 24s3 + 48s2 – 25s – 50 = 0
s0 –50 (s + 1)(s – 1)(s + j5)(s – j5)(s + 2) = 0

Week 4 69
Example 5: Control System Analysis

R K C
+ s( s  7)( s  11)
-

What range of K will cause the system to be stable, unstable, and marginally stable?
CLTF is:
C ( s) K
 3
R( s ) s  18s 2  77 s  K

Week 4 70
Routh’s Array
s3 1 77

s2 18 K
1386  K
s1
18
s0 K The sign depends on the value of K

For values of K:
• 0 < K < 1386 stable system
• K < 0 or K > 1386 unstable system

Week 4 71
Critical Stability
For K = 1386, there is an entire row of zeros.
By differentiating the auxiliary polynomial P(s) = 18s2 + 1386, we obtain:
dP( s )
 36s  0
ds
The Routh’s array is:
s3 1 77
s2 18 1386
s1 0 36
s0 1386

Week 4 72
Critical Stability (Cont’d.)
No sign change from the even polynomial (s2 row) down to the bottom, the even polynomial
has its two roots on the jw-axis of unit multiplicity.
Therefore, the system is marginally stable for K = 1386.

Week 4 73
Lesson Summary
At the end of this week, you should be able to:
• Identify poles and zeroes
• Assess system stability
Transient Steady-state
response response

y(t)
1

x(t)

0 10 20 30 Time (s)

Week 4 74