Chapter 6

Analysis of Feedback Control

Systems
Prof. Shi-Shang Jang
Chemical Engineering Department
National Tsing-Hua University
Hsin Chu, Taiwan
June, 2015
6-1 Introduction
Given a plant with a transfer function of
y(s)/m(s)=Gp(s), a controller is implemented
with a transfer function of Gc(s), the system
hence becomes a closed loop function.
It is very convenient to analyze the closed loop
system in s-domain since only algebraic
equations are involved.
In a closed loop system, there are two inputs for
the system, the set point (servo) of the output
and the disturbances (regulation) that affect the
output.
 6-1 Introduction – Cont.

Variations in outlet temperature are sensed by the sensor-transmitter and

sent to the controller causing the controller output signal to vary. This is
in turn causes the control valve position and consequently the steam flow
To vary. The variations in steam flow cause the outlet temperature to vary,
thus completing the loop.
Example: The Heat Exchanger System
6-1.1 Closed-Loop Transfer Function

E  s   K spToset  C  s 
M  s   Gc  s  E  s 
Ws  s   Gv  s  M  s 
To  s   Gs  s Ws  s   Gw  s W  s 
C  s   H  s To  s 
(1)Set Point Change
W s  0
To  s  K sp Gs  s  Gv  s  Gc  s 

To
set
s  1  H  s  Gs  s  Gv  s  Gc  s 
6-1.1 Closed-Loop Transfer Function -
Continued
(2)Load Changes
Toset  0
To  s  Gw ( s )

W s  1  H  s  Gs  s  Gv  s  Gc  s 
(3)General Case
 %TO   C   kg / s  %CO 
H  s  Gs  s  Gv  s  Gc  s       
 C   kg / s   %CO  %TO 
 dimensionless
Gw ( s ) K sp Gs  s  Gv  s  Gc  s 
To  s   W s   Toset  s 
1  H  s  Gs  s  Gv  s  Gc  s  1  H  s  Gs  s  Gv  s  Gc  s 

(4)Closed Loop Transfer Functions

Z

f Z=Output
Zi 1  l Zi=Input
f=product in forward path from Zi to Z
l=product of every transfer function in
feedback loop
6-1.1 Closed-Loop Transfer Function -
Continued

G1 s   Gv s Gs s H s 
G2 s   Gw s H s 
Gc s G1 s  G2 s 
C s   Rs   F s 
1  Gc s G1 s  1  Gc s G1 s 
Example
Example – Cont.
dT
Vc p  f t c pTi t   UATs t   T t  f t c pT t 
dt
dT
CM s  wt   UATs t   T t 
dt
Linearization
d
Vc p  c p Ti  T F t   UAs t    fc p  UAt 
dt
d
CM s  W t   UAs t   UAt 
dt
Laplacing
K K
s   F F s   s s s 
s  1 s  1
1 Kw
s s   s   W s 
 cs 1  cs 1
Example – Cont.

Control Valve Dynamics

Kv w ln  
Gv s   ; Kv 
 vs 1 100
Sensor  Transmitte r
KT 100  0
GT s   ; KT  1
T s 1 200  100
Example – Cont.
Example – Cont.
 Example – Cont.

KF Ks
s   F s   s
s  1 s  1
KF  Kw 1  Ks
 F s   W s   s  
s  1   cs 1  c s  1  s  1
 F s GF s   W s Gs s 
6-1.1 Closed-Loop Transfer Function -
Continued

G1 s   Gv s Gs s H s 
G2 s   Gw s H s 
Gc s G1 s  G2 s 
C s   Rs   F s 
1  Gc s G1 s  1  Gc s G1 s 
Example – Cont.

1.652 1.183 1.0

G1  s   Gv  s  Gs  s  H  s  
0.2 s  1 8.34 s  1 0.502 s  1 0.75s  1
3.34  0.524 s  1 1.0
G2  s   GF  s  H  s  
8.34s  10.502s  1 0.75s  1
Gc  s  G1  s  G2  s 
C s  R s   F s 
1  Gc  s  G1  s  1  Gc  s  G1  s 
6.1.3 Steady State of Closed Loop Systems

To s  K sp Gc s Gv s Gs s  To s  Gw s 
 
To s  1  H s Gc s Gv s Gs s 
set W s  1  H s Gc s Gv s Gs s 
To 0  Gw 0 
To 0  K sp Gc 0 Gv 0 Gs 0  
W 0  1  H 0 Gc 0 Gv 0 Gs 0 

Toset 0  1  H 0 Gc 0 Gv 0 Gs 0  
Kw

Kw
1  K T K c K v K s 1  K OL
K sp K c K v K s K sp K c K v K s
 
1  KT K c K v K s 1  K OL
Example
Example: New steady state of the heating tank

1. Servo control with Kc=1;Tset=1

T Kc Kv K s 2 1.652  1.183 3.9086
    0.7963
Tset 1  K c K v K s 1  2 1.652  1.183 4.9086
Tset  1
T  150  1 T  150.7963

2. Regulation Control with F=1

T KF 3.34 -3.34
    -0.6804
F 1  K c K v K s 1  2 1.652 1.183 4.9086
T  150  T  149.3196
Example: Flow rate control system
Example: Flow rate control system –
Cont.
F s  Kv
Gv s    ; H s   K T
M s   v s  1
F s  K sp Gv s Gc s 

F s  1  K T Gv s Gc s 
set

 1 
Gc s   K c 1  
 Is 
F s  K T K v K c  I s  1

F set s   I s  v s  1  K T K v K c  I s  1
Set  I   v
F s  KT K v K c 1
 
F set s   v s  K T K v K c  FC s  1
6.1.2 Characteristic Equation of the
Loop
  v  c 
1  H ( s )G s G s G s  0
s

an s n  an1s n1    a0  0
an  s  r1  s  r2   s  rn   0
Gw  s 
To ( s )  W s 
1  H ( s )Gs  s  Gv  s  Gc  s 
numerical terms
To ( s ) 
an  s  r1  s  r2   s  rn  input terms 
b1 b b
To t    2    n  input terms
s  r1 s  r2 s  rn
 Unforced terms  Forced terms
Example : Servo Problem for a First Order System
K
G1 s   ; F s   0
s  1
Case 1 : Gc  K c ; Proportional only
K Kc K
Kc
C s  s  1  Kc K 1  Kc K
 
Rs  1  K K s  1  K c K s 1
1  Kc K
c
s  1
K'

 ' s 1
 1 
Case 2 : Gc  K c 1  ; (PI)
  s
I 

 1  K K c  I s  1 K
K c 1  
C s    I s  s  1  Is s  1

Rs   1  K K c  I s  1 K
1  K c 1   1
  I s  s  1  I s  s  1
KK c  I s  1

KK c  I s  1   I s s  1
6.1.3 Steady State of Closed Loop Systems

 Proposition 6-1 ： The closed loop dynamics of a first

order system with a P-only controller behaves the same
with another first order system, with a gain of KOL/(1+
KOL) and time constant of  /(1+ KOL) for the servo
problem.
6.1.3 Steady State of Closed Loop Systems–
Cont. 
Corollary 6-1 ： A P-only controller applied to a first order system
can yield perfect control if Kc for a servo problem.
 Proof: Consider (6-1), in case , we have K11, and 10.
 Corollary 6-2 ： Assume that a step change with a magnitude of A
at the set point Tset(s) is implemented to a P-only controller applied
to a first order system, then the offset of the controller is A/(1+KOL)
.
 Proof: Let Tset=A/s , then To(s)=(A/s)[K1/ (1s+1)] , we have:
AK1 AK OL
lim To t   lim sTo s   lim  AK1 
t  s 0 s 0  s  1 1  K OL
1

1
A  AK1  A
offset = 1  K OL
6.1.3 Steady State of Closed Loop Systems–
Cont.
1

Constant simout

To Workspace

2
2
5s+1
Add Gain Scope
Transfer Fcn

Kc 1 K1

1 1.67 0.667
2 1 0.8
5 5/11 10/11
 6.1.3 Steady State of Closed Loop Systems–
Cont.
Regulation problem (disturbance rejection) using a
P-only control for a first order system:
K K2
G1  ; G2  ; Rs   0
s  1  2s 1
K2
C (s) G2 s   2s 1
 
F s  1  Gc s G1 s  1  K K
c
s  1
C ( 0) K2

F 0  1  K c K1
K2
Offset   F 0 
1  K c K1
 6.1.3 Steady State of Closed Loop Systems–
Cont.
Corollary 6-3 ： A P-only controller
applied to a first order system can yield
perfect control if Kc for a regulation
problem.
Corollary 6-4 ： Assume that a step
change with a magnitude of A at the load
F(s) is implemented to a P-only controller
applied to a first order system, then the
K
offset of the controller is 1  K K F 0.
2

c

1
6.1.3 Steady State of Closed Loop Systems–
Cont.

Kc=1

Kc=2

Kc Offset
Kc=5

1 -0.124

2 -0.077

5 -0.036

time
 6.1.3 Steady State of Closed Loop Systems–
Cont.
Proposition 6-2 ： The offset of closed loop
dynamics of a first order system with a PI
controller results zero offset.
Proof:

C s  KK c  I s  1

Rs  KK c  I s  1   I s s  1
C 0  KK c
 1
R0  KK c
6-2 Stability of the Control Loop

Let r1 , r2 ,, rn be n roots of 1  Gc  s  G1  s   0

b1 b b
C t    2    n  input terms
s  r1 s  r2 s  rn
=b1e r1t  b2 e r2t    bn e rnt  particular solution
At the cross over point:
b s  b2
C  s   12   other terms 
s  u 2

 C t   b1 'sin u t   
6-2.1 Stability Criterion
Proposition 6-3: For a feedback control loop to
be stable, all of the roots of its characteristic
equation must be either negative real number or
complex numbers with negative real parts.
Corollary 6-4: For a feedback control loop to
be stable , all of the roots of its characteristic
equation must fall on the left-hand half of the s
plane, also known as the “left-hand plane”
(LHP).
 6-2.1 Stability Criterion -
Example
1
G1  s  GP  s  Gv  s  H  s  
 s  1 s  2  s  3
1
KC
C s  s  1 s  2  s  3
Kc p1 p2 p3

R s 1 K
0 -3 -2 -1
1
C
 s  1 s  2  s  3 0.23 -3.1 -1.75 -1.15

Or 0.39 -3.16 -1.42* -1.42*

C s KC
1.58 -3.45 -1.28-0.75j -1.28+0.75j

R  s   s  1 s  2  s  3  K C 6.6 -4.11 -0.95-1.75j -0.95+1.75j

26.5 -5.1 -0.45-2.5j -0.45+2.5j

Characteristic equation can be derived: 60* -6.0 0-3.32j* 0+3.32j*

s  1s  2s  3  K C 0
100 -6.72 0.35-4j 0.35+4j
6-2.1 Stability Criterion - Example
6-2.1 Stability Criterion – Cont.
 Consider the previous example, at Kc=0.39, the system
becomes oscillatory, at Kc=60, the system becomes
unstable. At this point, we term the controller gain that
makes the closed loop system unstable the ultimate gain
(Ku=60) of the system, the frequency (u=3.32) that the
system oscillates is termed the ultimate frequency.

 Problem: Given the characteristic equation C(s, Kc)=0,

what is the ultimate gain and ultimate frequency of the
system?

 Answer: Direct substitution, Routh array, Root Locus

6-2.2 Method of Direct Substitution – Cont.
Basic idea: Consider the characteristic equation
C(Kc,s)=0, in case of ultimate gain, i.e. Kc=Ku,
then the root of the equation should be located at
the imaginary axis, i.e. s=j. The problem of
solving the ultimate condition, hence changed
into C(Kc,s)=R(Kc, )+jI(Kc, )=0, i.e. R(Kc,
)=0, I(Kc, )=0 (two unkowns, two equations)
Problem: many algebraic work invloved.
6-2.2 Method of Direct Substitution – Cont.

K u  s  1s  2 s  3  K u  s 3  6 s 2  11s  6

 K u  ju  6 ju  11 ju   6
   
3 2

   ju  11 ju  j 6  K u  6u 

3 2

0
11u  11u  0; 6  K u  6u  0
3 2

u  11; K u  60
6-2.2 Method of Direct Substitution – Cont.

Fig. 6  1.3
50 C 1 % 0.016 kg / s
Gs s   ; H s   ; Gv s  
30s  1 kg / s 10s  1 C 3s  1 %CO
50 1 0.016
1  Gs s H s Gv s Gc s   Kc  0
30s  1 10s  1 3s  1
30s  110s  13s  1  0.8K c  0
900s 3  420s 2  43s  1  0.8 K c  0
Let s  ju
 420 2
u   
 1  0.8 K c  j  900u3  43u  0
u  0.2186rad / s; K cu  23.8
2
Tu   28.7
0.2186
6-2.3 Effect of Parameters
Let the temperature sensor/transmiter is re-calibrated (range reduced to
75-125C such that
2 %
H s 
10 s  1 C
50 2 0.016
1  Gs  s  H  s  Gv  s  Gc  s   1  Kc  0
30 s  110 s  1 3s  1
900 s 3  420 s 2  43s  1  1.6 K c  0
2
u  0.2186rad / s; K cu  11.9; Tu   28.7
0.2186
Let the temperature sensor/transmiter is replaced to a faster one such that:
1 %
H s 
5s  1 C
50 1 0.016
1  Gs  s  H  s  Gv  s  Gc  s   Kc  0
30 s  1 5s  1 3s  1
2
u  0.2345rad / s; K cu  18.7; Tu   26.8
0.2186
6-2.3 Effect of Parameters – Cont.
6-2.4 Effect of Dead Time – Pade
Approximation

2 2 3 3
t 0 s t s t s
e  1  t0 s  0
 0
 ...
2 6
t0 s
1 2 2 3 3
t0 s t0 s
 2  1  t0 s    ...
t0 s 2 4
1
2
6-2.4 Effect of Dead Time – Pade
Approximation – Cont.
Ke  t0 s
G1 s  
s  1
K c Ke t0 s
1  K c G1 s   1  0
s  1
K c K 1  0 
ts
 2
1 0
1  0 s  1
t s
 2 

t0 2  t t 
s    0  K c K 0  s  1  K c K  0
2  2 2
2 t0
u  1
t0 
 
KK cu  1  2 
 t0 
Homework
Text , page 220
6-1, 6-3, 6-5, 6-10, 6-12, 6-18, 6-20, 6-25
 Example
 The temperature of a tank is controlled by adjusting the steam
flow to the jacket of the tank. The temperature transmitter has
a span of 100°F and is set between 100 and 200°F. The
proportional controller is set at a proportional band of 20. The
normal pressure to the valve is 8 psig and the valve is air to
open type. The normal temperature of the tank is 170°F and
the normal feed temperature is 65°F. The set point is moved
from 170°F to 175°F. And the tank eventually comes to a
steady state of 174.1°F.
◦ (i) What is the offset?
◦ (ii) What is the pressure to the valve in the final steady state?
◦ (iii) What is the process gain? (Assume the valve gain is unity)
◦ (iv) What will be the offset if the proportional band was changed to 10?
Solution
(i )175  174.1  0.9 F
0.9
(ii )e  12  0.108; Kc  100 / 20  5;
100
M  5  0.108  8  8.54 psig
174.1  170
(iii ) K p   7.59
5  0.108
Y 10  7.59
(iv)   0.9
Yset 1  10  7.59  0.12
T  170  5  0.9  174.5 F