CN3121 Process Dynamics and Control

10. Dynamic Behavior of Closed-loop Control Systems

Min-Sen Chiu Department of Chemical and Biomolecular Engineering National University of Singapore

Dynamic Behavior of Closed-Loop 10 Control Systems

Become familiar with the major elements in the feedback control system Develop closed-loop transfer functions Evaluate the dynamic behavior of processes operated under feedback control

10
Dynamics of Closed-Loop Control Systems

The combination of the process and the feedback controller is called the closed-loop system. Variables of a closed-loop system: 1. Inputs set-point and disturbance variables. 2. Output - controlled variable. The analysis of closed-loop systems can be difficult due to the presence of feedback. Two useful tools: 1. Block diagram 2. Closed-loop transfer function Block diagrams can provide quantitative information if each block is represented by a transfer function.
3

Development of a Block Diagram

10
Dynamics of Closed-Loop Control Systems

Return to the previous example of stirred tank blending process (see Chapters 2 and 4, SEM).

Control objective: Regulate tank composition x Manipulated variable: Flow rate of pure A, w2 Primary disturbance: Inlet composition x1 Assume w1 constant Derive transfer functions for each component in the closed-loop system.
4

Process

10
Dynamics of Closed-Loop Control Systems

Approximate dynamic model of the stirred-tank blending system is available (see eq. 4-69, SEM):

K K X ( s ) = 1 X1 ( s ) + 2 W2 ( s ) s + 1 s + 1
where =

(11-1)

V , w

K1 =

w1 , and w

K2 =

1 x w

(11-2)

Composition Sensor

10
Dynamics of Closed-Loop Control Systems

Assume a first-order transfer function:

Xm (s) Km = X ( s ) m s + 1

(11-3)

Usually m << . Compared to the (slow) process dynamics, sensor dynamics is considered as negligible (fast) dynamics; thus its transfer function can be further simplified as a steady-state gain Km.
6

10
Dynamics of Closed-Loop Control Systems

Controller Suppose that a proportional plus integral (PI) controller is used. The controller transfer function is

P ( s )

1 = Kc 1 + E (s) I s
and the error signal e(t)

(11-4)

P ( s ), E(s) - Laplace transforms of the controller output p ( t )

p and e - electrical signals (units of mA)

Set point expressed as the actual physical variable

Set point expressed as an electrical current signal

The error signal is

' e(t ) = ~sp (t ) xm (t ) x'

(11-5)

10
Dynamics of Closed-Loop Control Systems

Transform

~' ' E ( s) = X sp ( s) X m ( s)
set-point x'sp (t ) by sensor gain Km :

(11-6)

~' (t) is internal set-point related to the actual composition x sp

~ ' (t ) = K x ' (t ) xsp m sp

(11-7)

~' X sp ( s ) X (s)
' sp

= Km

(11-8)

Eqs. 11-4, 11-6 and 11-8 are shown in the controller block diagram.

10
Dynamics of Closed-Loop Control Systems

Current-to-Pressure (I/P) Transducer Usually has linear characteristics and negligible (fast) dynamics; thus assume that the transfer function merely consists of a steady-state gain KIP.

Pt( s ) = K IP P ( s )

(11-9)

Control Valve

10
Dynamics of Closed-Loop Control Systems

Assume that the valve can be modeled as

W2 ( s ) Kv = Pt( s ) v s + 1

(11-10)

[psi]

The valve dynamics is generally nonlinear => approximated by linear (1st order) model in the vicinity of the nominal operating condition
10

10
Dynamics of Closed-Loop Control Systems

Combining the block diagrams for the individual components, we get the composite block diagram of the control system:

11

Block Diagram Simplification

10
Dynamics of Closed-Loop Control Systems

Basic elements in a block diagram:

c
p

Circle represents algebraic relation of the input arrows, e.g. e = r c. Arrow indicates flow of information, e.g. p = Ge = G(r - c). Block represents the relevant dynamics (by transfer function model) between the input and output.
12

Some basic rules

10
Dynamics of Closed-Loop Control Systems

1. Y = A B C
A + B
A-B

+ C

-B-C +

Y +

2. Y = G1G2A

G1

G2

A A

G2

G1 Y

G1G2

13

10
Dynamics of Closed-Loop Control Systems

3. Y = G1(A B)
A + B B G1 Y A G1 G1 + Y

4. Y = (G1+G2)A
G2A G1A

G1 G2

+ +

G2

G1/G2

+ +

G1+G2

Y
14

Block Diagrams General Treatment

10
Dynamics of Closed-Loop Control Systems

This is a general diagram that can be used to represent a wide variety of practical control problems.

Gp Effect of manipulated variable on the controlled variable Gd Effect of load variable on the controlled variable
15

Note

10
Dynamics of Closed-Loop Control Systems

Each variable in the figure is the Laplace transform of a deviation variable. For simplicity, the primes and s have been omitted; thus Y means Y(s).

Standard symbols G = transfer function (subscripts c, v, p, d, m = controller, valve, process, disturbance, and measurement respectively) Y - process output Ym - measured output
sp

D - disturbance or load variable Ysp - set-point E - error U - manipulated variable

- internal set-point

P - controller output

Closed-Loop Transfer Functions

10
Dynamics of Closed-Loop Control Systems

The objective is to find the transfer functions between the inputs (Ysp and D) and the output (Y) of the closedloop system. For the process input,

~ U = G v P = G v G c E = G v G c (Ysp Ym ) = G v G c ( K mYsp G mY )
Process output is obtained as

Y = G pU + G d D
From the above two equations,

Y = G p G v G c ( K mYsp G m Y ) + G d D
17

10
Dynamics of Closed-Loop Control Systems

Rearranging,

Y =

G p GvGc K m 1 + G p GvGcGm
Effect of Ysp on Y

Ysp +

Gd D 1 + G pGvGcGm
Effect of D on Y

(11-30)

Eq. 11-30 illustrates the important role of Laplace Transform in analysis of feedback control system

18

Two Types of Control Problem

10
Dynamics of Closed-Loop Control Systems

1. Servo problem (set-point change) Assume Ysp 0 and D = 0 (set-point change while disturbance change is zero). From the last equation,

G pGvGc K m Y = Ysp 1 + G p G v G c G m

(11-26)

(closed-loop transfer function for set point change)

2. Regulator problem (disturbance change) Assume D 0 and Ysp = 0 (constant set-point)

Gd Y = D 1 + G p GvGcG m

(11-29)
19

(closed-loop transfer function for load change)

Remarks

10
Dynamics of Closed-Loop Control Systems

1. Closed-loop transfer functions (eqs. 11-26 and 11-29) depend on dynamics of process, measurement device, controller and control valve. 2. Denominator for both transfer functions, eqs. 11-26 and 11-29, is the same => (1 + product of all the transfer functions in the loop), i.e. (1+GpGvGcGm). 3. Overall transfer function = (product of transfer functions in the forward path)/(1 + product of all transfer functions in the loop). 4. (1+ GpGvGcGm) is often written as (1+ GOL) where GOL= GpGvGcGm is the open-loop transfer function. GOL relates Ym to sp if the feedback loop is opened just before the comparator. 5. For simultaneous changes in disturbance and set-point (i.e., D 0 and Ysp 0), eq 11-30 holds => overall response is the sum of the individual responses. 20

10

Remark 3

10
Dynamics of Closed-Loop Control Systems

Negative feedback

G cGvG p K m Y = Y sp 1 + G cG vG pG m
Gd Y = D 1 + G cG vG pG m

Forward path from Ysp to Y

Forward path from D to Y Product of all transfer functions in the loop

Consider this feedback system,

10
Dynamics of Closed-Loop Control Systems

Ysp

Km

Gc

Gv Gm

Gp

Negative feedback

G cGvG p K m Y = Y sp 1 + G cG vG pG m
Gp Y = D 1 + G cG vG pG m

Forward path from Ysp to Y

Forward path from D to Y Product of all transfer functions in the loop

11

Feedback system equivalent diagrams

10
Dynamics of Closed-Loop Control Systems

D Ysp Km

~ +Ysp
-

Gc

Gv Gm

Gp

~ Ysp
Ysp Km + Gc Gv Gm
Gp D 1+GcGvGpGm + Ysp GcGvGpKm 1+GcGvGpGm + Y

Gp Gp + + Y

23

Analysis and Design Problems

10
Dynamics of Closed-Loop Control Systems

Analysis Given particular Gp, Gv, Gc, Gm,

- Is the closed-loop system stable? - Speed of response? Damping?

Design Given particular Gp, Gv, Gm, and Gd, design Gc so that - The closed-loop dynamics are stable - Y/Ysp has a gain of ?? and Y/D has a gain of ?? - The dynamics are sufficiently fast and smooth (without excessive oscillations)

24

12

Effect of Proportional Control on Closed-Loop Response

10
Dynamics of Closed-Loop Control Systems

Consider 1st-order process and P-controller. Then

Gp =

Kp

s + 1

Gd =

Kd s + 1

Gc = K c

For simplicity, let Gv = K v and Gm = K m

25

From eq. 11-30,

10
Dynamics of Closed-Loop Control Systems

Y =

KcKvK pKm

s + 1 + K c K v K p K m
K1 K2 Ysp + D 1s + 1 1s + 1

Y sp +

Kd D s + 1 + K c K v K p K m
Note: CL system is 1st order with time const 1. For both TFs, is the same, but gain is different

Decreases with increasing Kc Always < , i.e., CL response is faster than OL response 1 unless Kc = Always < 1 0 unless Kc = Always < Kd

1 =

1 + Kc Kv K p Km

K1 =
K2 =

Kc Kv K p Km 1+ Kc Kv K p Km
Kd 1 + Kc Kv K p Km
26

13

Servo problem

10
Dynamics of Closed-Loop Control Systems

Step change of magnitude M in set-point, i.e., Ysp = M/s and D = 0.

K1 M 1s + 1 s Inverting, y(t ) = K1M (1 e t /1 )

Y =

M 1+ Kc Kv K p K m

Less than the desired value M.

27

10
Dynamics of Closed-Loop Control Systems

As t , output response never reaches new set-point. The discrepancy is called (steady-state) offset. Offset = current set-point (final value of the response) = M K1M = M

MK c K v K p K m 1+ Kc Kv K p Km

M 1 + Kc Kv K p Km

Offset decreases with increasing Kc. This is the characteristic of P-control. Theoretically, offset 0 when Kc . But, does it happen? If not, why?

28

14

Regulator problem

10
Dynamics of Closed-Loop Control Systems

In this case, D = M/s Ysp = 0 and Y = Inverting, y (t ) = K 2 M (1 e t /1 )

K2 D 1s + 1

D(t) no control (Kc=0) with control

y(t)
K2M

0 0 Time

Offset = current set-point (final value of the response) KdM = 0 K 2M = 1+ KcKvK pKm Kc offset
29

Proportional-Integral Control for Disturbance Change

10
Dynamics of Closed-Loop Control Systems

In this case, Gc = K c (1 +

Y = D

Kd K d I s s + 1 = K KK K 1 I s(s + 1) + K c K v K p K m ( I s + 1) 1 + c v p m (1 + ) s + 1 Is

1 ) Is

Rearrange

Y K 3s = 2 D 3 s + 2 3 3 s + 1

Gain = ?

where

K3 =

Kd I 1 ( Kc Kv K p Km +1) I I , 3 = , 3 = K c Kv K p K m Kc K v K p K m 2 Kc Kv K p K m
30

15

For a unit step change in load, D = 1/s and hence

10
Dynamics of Closed-Loop Control Systems

Y (s) =

K3 s + 2 3 3 s + 1
2 3 2

The output response for 3 < 1 is

1 y (t ) = K 3 3

1 1 32

3t / 3

t sin 1 3
2 3

For this specific case, the responses for different Kc and I are given in the next slide.

31

= 1, Kp = 1, Kv = 1

10
y(t)
Dynamics of Closed-Loop Control Systems

I = 0.25

y(t)

Kc = 3.5

Remarks 1. Integral action eliminates offset, ysp() y() = 0 2. For Kc or I 3. For Kc 4. For I response speeds up response more oscillatory (unexpected) response more oscillatory (expected)

5. Note In general, closed loop response becomes more oscillatory as Kc. The anomalous result above is due to neglected valve and measurement dynamics. When these are included, the TF is 32 no longer 2nd-order.

16

Proportional-Integral Control for Set Point Change

10
Dynamics of Closed-Loop Control Systems

For this case,

K c K p (1 + 1 / I s ) /(s + 1) Y = Ysp 1 + K c K p (1 + 1 / I s ) /(s + 1)

Standard form

s +1 Y = 2 2 I Ysp 3 s + 2 3 3 s + 1
Introduce unit step change in Ysp and inverse

Gain = 1 always! No offset 2nd order dynamics 3, 3 as defined earlier

y (t ) =

I 3 1
2

e t / 3 sin 1

+1

1 1
2

1 2 t e t / 3 sin 1 2 + tan 1 3 33

10
y(t) Kc = 1, Kp = 1 I = 1, = 1
Dynamics of Closed-Loop Control Systems

Again, zero offset Offset = ysp() y() = 1 1 = 0

34

17

Effect of Measurement Lag

10
Dynamics of Closed-Loop Control Systems

As before let

Gp =

Kp

s + 1

Gd =

Kd , s + 1

Gc = K c , Gv = K v = 1
1 ms +1

Assume significant measurement lag: G m = Resulting control system is

Gd =
Gp =

Kd s + 1
Kp
+ + Y

Ysp

Kc

s + 1

Gm =

1 m s +1

35

Consider set point change,

10
Dynamics of Closed-Loop Control Systems

KcK p Y = Ysp

s + 1
1+ KcK p (s + 1)( m s + 1)

K 5 ( m s + 1) s 2 + 2 5 5 s + 1
2 5

where K 5 =

Kc m , 5 = , 1+ K p Kc 1 + K p Kc

and

5 =

+ m 2 m

1 1+ K p Kc

Produces 2nd-order system even for P-control. Response may be oscillatory depending on the choice of , m, Kp and Kc. One possibility is as follows:
36

18

10
y(t)
Dynamics of Closed-Loop Control Systems

=1

Kp = 1

Kc = 8 Ysp = 1/s

Remarks Measurement lag produces poorer transients Offset results from the use of P controller
37

Control of Pure Capacity (or integrating) Process

10
Dynamics of Closed-Loop Control Systems

(const)

h controlled by manipulating q1. Mass balance (in terms of deviation variables):

dh = q1 dt

or

h=

q1 As

38

19

Let Gc = Kc, Gv = Gm = 1.

10
hsp Kc
Dynamics of Closed-Loop Control Systems

q1

1/As

h=

1 hsp ( As / K c ) + 1

For set-point change hsp = 1/s, h(t) = lim [s h(s)] = 1

h=

1 1 ( As / K c ) + 1 s

Therefore offset = hsp h (t) = 1 1 = 0 Alternatively, ..

Summary

10
Dynamics of Closed-Loop Control Systems

When we place chemical process in a closed-loop with sensors, transducers, valves and controllers, we have a more complicated system than the original process. Nevertheless, our analysis shows that a closed-loop system can be written as one single transfer function. Thus we recognize that a block diagram provides a convenient representation for analyzing control systems. We understand the key features of P and PI control using simple first-order processes. We notice that measurement dynamics can cause deterioration in the control system performance.

Further reading:

Chapter 11.1 and 11.3, SEM

20