Sunteți pe pagina 1din 60

Outline

 Part 1: Continuous SISO systems


 Introduction
 Control Objectives in the Presence of Uncertainty
 Modeling Uncertainty
 Nominal Stability & Performance
 Robust Stability
 Robust Performance
 IMC Structure
 Stability and Performance
 Prefect Control
 IMC Design Procedure
 Stable systems
 Example1: System with time delay uncertainty

2
Outline
 IMC based PID
 Introduction
 General Relationships
 Example 2: PID Design
 Example 3: Model with high uncertainty
 Part 2: Continuous MIMO systems
 IMC Structure - MIMO Case
 Internal stability for stable plants
 General Internal stability
 IMC Design Procedure - MIMO Case
 Inner-Outer factorization
 Nominal Performance
 Robust Stability & Performance

 Exercises
3
 Part 1: Continuous SISO systems
 Introduction
 Control Objectives in the Presence of Uncertainty
 Modeling Uncertainty
 Nominal Stability & Performance
 Robust Stability
 Robust Performance
 IMC Structure
 Stability and Performance
 Prefect Control
 IMC Design Procedure
 Stable systems
 Example1: System with time delay uncertainty

4
Introduction
 IMC is an effective method for designing and implementing
robust controllers.
 IMC structure is an alternative to the classic feedback structure.
 Its main advantages are:
 Simple and easy to understand designing procedure
 On-line tuning of IMC controllers are very convenient
 It can easily control plants in presence of actuator
constraints.
 use uncertainty information in design procedure.

5
 Part 1: Continuous SISO systems
 Introduction
 Control Objectives in the Presence of Uncertainty
 Modeling Uncertainty
 Nominal Stability & Performance
 Robust Stability
 Robust Performance
 IMC Structure
 Stability and Performance
 Prefect Control
 IMC Design Procedure
 Stable systems
 Example1: System with time delay uncertainty

6
lm (w)

 p(iw)  ~p (iw)  la (iw)  p(iw)  ~p (iw)(1  lm (iw))


 
 la (iw)  la ( w)  lm (iw)  lm ( w)

  p : p(iw)  ~
p (iw)  la ( w)

Uncertainty usually increases with frequency.


Nominal Actual
model Plant 7
la (w)

Controller Model
Control Objectives
 Nominal Stability: The system is stabe with no model
uncertainty.
 Nominal Performance: The system satisfies the performance
specifications with no model uncertainty.

 Robust Stability: The system is stabe for all perturbed plants,


here means for all family  plants.
 Robust Performance: The system satisfies the performance
specifications for all family  plants.

8
Remember: Nominal Performance d

y ( s )  T ( s )r  S ( s )d  T ( s )n r
c
~
p
y

the sensitivity function S is a very good indicator of closed-loop


performance
1
 S ( j )   ,   wP ( j ) S ( j )   1
wP ( j )

H  optimal controller is obtained by soliving : min N ( K ) 


K
9
la (w)

Controller Model
Robust Stability
Theorem 1 : Assume that controller c stabilizes the
nominal plant ~
p . Then the system is stable for family
 p ~ p (iw) 
  p : ~  lm ( w)
 p 
~ ~
If and only if T lm  sup T lm  1

w
~
Model uncertainty imposes a bound on T.
~
Noise also tends to impose a bound on magnitude of T
but usually the constraint imposed by model uncertainty
tends to dominate.

10
Robustness / Performance Tradeoff
~ ~
T lm  sup T lm  1

w

y ( s )  T ( s )r  S ( s )d  T ( s )n
~ ~
we want to minimize S w p for nominal performance and T lm
 
for robustness. This problem ic inherent in feedback control and
cannot be removed by clever design.

11
la (w)

Controller Model
ROBUST PERFORMANCE
Theorem 2 : Assume that controller c stabilizes the
nominal plant ~
p, then the closed - loop system will meet
the performance specification wP ( j ) S ( j )   1 p  
~ ~
if and only if T lm  S w p  1 w
In practice our objective is to find a controller which solve
min sup ~ ~
T lm  S w p 
c w

It is a difficult problem to solve. For general MIMO case no


reliable solution are available
IMC easily provides a good approximation to the optimal
solution 12
 Part 1: Continuous SISO systems
 Introduction
 Control Objectives in the Presence of Uncertainty
 Modeling Uncertainty
 Nominal Stability & Performance
 Robust Stability
 Robust Performance
 IMC Structure
 Stability and Performance
 Prefect Control
 IMC Design Procedure
 Stable systems
 Example1: System with time delay uncertainty

13
IMC Structure
d
r y
Controller Plant

Model

Because in addition to the controller, It includes the plant model


explicitly this feedback configuration is called
internal model control (IMC)
~
The feedback signal d expresses the uncertainty
(model uncertainty and disturbance).
14
IMC vs. Classic Feedback

q c
c ~ , q
1  pq 1 ~
pc

So Why IMC has so much advantages over classic feedback?!!

15
~p

Stabilizing Controllers c  q
1 ~
pq

Assume plant is stable, so the only requirement for nominal stability is stability of q
~
pq 1 ~pq S  1  ~
pq
y ~ r  ~ d ~ ~
1  q( p  p ) 1  q( p  p )
T  pq
~ ~
pc
T 
1 ~
pc
16
lm (w)

Prefect Control
pq 1 ~pq
y ~ r  ~ d
1  q( p  p ) 1  q( p  p )
If we put q  1 ~ then T  1 and S  0, so we can achieve
p
prefect control.
But there are three reasons that make the prefect control impossible:
1) Nonminimum-phase(NMP) plants: q become noncasual or
unstable, and nominal plant become unstable.
2) Strictly proper plants: q become improper, but this can be solved
by adding some far poles.
3) Model Uncertainty:
~ ~
T lm 1  (1  S )lm 1
 
~
S 0  lm  1
17
 Part 1: Continuous SISO systems
 Introduction
 Control Objectives in the Presence of Uncertainty
 Modeling Uncertainty
 Nominal Stability & Performance
 Robust Stability
 Robust Performance
 IMC Structure
 Stability and Performance
 Prefect Control
 IMC Design Procedure
 Stable systems
 Example1: System with time delay uncertainty

18
IMC Design For Stable Systems

 Final Objective : min sup T~lm  S~w p 


c w
IMC design procedure consists of two steps:

Step 1: Nominal Performance

q~ is selected to yield a good performance without regard


for constraints and model uncertainty
~
min e  min S w  min (1  ~
2
pq~ ) w 2
q~ q~ 2 q~

Step 2: Robust Stability and Performance

19
Step 1:Nominal Performance
Theorem3 (H 2 optimal control) : assume ~ p is stable,
factor ~
p and input w into an allpass portion ~
p and a MP
A
portion ~pM , ~
p~pA  ~ pM .
so that ~
pA includes all the RHP zeros and delays of ~
p and
~
p (iw)  1 w , The controller ~ q which solves above equation is :
A
~
q  (~ 
pM wm ) 1 ~
p A1wm *
Where the operator 
.  denotes that after a partial fraction expansion of
p A1 are omitted.
the operand all terms involving the poles of ~

20
~ ~ s2 s 1
Example : p  p A  , vm 
s2 s
~ s  s  2 s  1 s  1 6   s 1
q     1     .
s  1   s  2 s  s  1  s  s  2  s  1

 In general the optimal controller ~


q in not proper.

~
min S w
q~ 2

~ ~
T lm  sup T lm  1

w
21
Step2: Robust Stability
and Performance
q  q~f

~
S  1  ~
pq ~ ~
pc
~ ~ T 
T  pq 1 ~
pc

For robust stability ~q is augmented by a low - pass filter f


to simplify t he design tas k.
Low-pass filter can improve robustness.
What about performance?

22
we fix the filter structure and search over a small number
m1 1
of filter parameters, f  (  m1s  ...1s  1) .
(s  1) n
Here n is selected large enough for q to be proper.To have
zero steady state error, the condition on f to be satisfied are:
1
Type1 : f (0)  1  f ( s ) 
(s  1) n
df ns  1
Type 2 : f (0)  1 and lim  0  f (s) 
s 0 ds (s  1) n
 is the adjustable filter parameters, Increasing  will
increase robustness but it decrease the performace.
IMC makes the on - line tuning so easy and intuitive. 23
Example 1: System with time delay uncertainty
-s -2s
2e 2e
Plant : ~
p , Model : ~
p
0.5s  1 0.5s  1
Input: Step
Uncertainty: Time delay uncertainty
lm (iw)  e iw  1
  2   ,   0.5 so : 
   0.5 lm ( w)  e iw0.5  1 w  2
Step1 : Nominal Performance : lm ( w)  2 w  2
p  e 2 s
~
A p 1  0.5(0.5s  1)
, q~  ~ m

Step2 : Robust stability and Performance :


1
f 
s  1 24
~ ~
T lm  sup T lm  1
Example 1 
w

1.4
  0.2 2
1.2
1.5
1

0.8

lm
1
Tlm

0.6

0.4 0.5

0.2
0
-1 0 1 2
0 10 10 10 10
-1 0 1 2
10 10 10 10 w

  0.6
0.8

0.6

0.4

0.2

0
-1 0 1 2
10 10 10 10 25
2e-2.45s
assume real plant is : p 
Example 1 0.5s  1
  0.2   0.6
Step Responselambda 0.2
30
Step Response Step Responce, Lambda 0.5
1.4
Step Response
20
1.2

10 1

0.8
Amplitude

Amplitude
0.6
-10
0.4

-20
0.2

-30 0
0 10 20 30 40 50 60 0 5 10 15 20 25 30
Time (sec) Time (sec)

q 0.5(0.5s  1)
c  How to implement?!!!!
1  pq 0.6s  1  e 2 s
~

26
 IMC based PID
 Introduction
 General Relationships

 Example 2: PID Design

 Example 3: Model with high uncertainty


 Part 2: Continuous MIMO systems
 IMC Structure - MIMO Case
 Internal stability for stable plants
 General Internal stability

 IMC Design Procedure - MIMO Case


 Inner-Outer factorization

 Nominal Performance

 Robust Stability & Performance

 Exercises
27
Introduction

 By far the most widely used controllers in the industry


processes are the two-term PI and the three-term PID
controller.

 Because IMC is clearly more general and therefore more


powerful it is worthwhile to explore the relationships between
IMC, PI and PID in order to gain insight into the tuning of
these simple controllers, their performance and robustness.

28
General relationships
 Remember: General relationship between the classic feedback
controller c and the IMC controller q are:

q c
c , q
1  pq 1  pc
f the plant m odel . So The com plexity of the equivalent classic controller c is
 The order of q obtained in this way is generally higher than the
order of the plant model. So The complexity of the equivalent
classic controller c is determined by the complexity of the
model.

Simple models simple controllers

29
Example 2: PID Design

 In the previous example , our model was as fellows:


~ 2e 2 s
p
0.5s  1
 By the use of “Pade” approximation our model is obtained as
follows:
~ 1 s 2
p .
1  s (.05s  1)
~
pA ~
pM
 Step 1: Nominal Performance

q~  ( ~ p A1 wm *
pM wm ) 1 ~
30
Example 2
1
 2 1  1  s 1  0.5s  1
q  .   .  
 .05s  1 s  1  s s * 2

Step 2: Robust stability and performance:

0.5s  1 1
q  qf q .
2 s  1
So:
0.25s  0.5
q s  1 (s  1)(0.25s  0.5)
c  
1  pq 1  1 s 0.25s  0.5 s ( s  2   )
.
(.05s  1)(s  1) s  1
31
Example 2
Structure of the PID controller is defined as follows:
kI 1
K (s )  (k P   k D s ).
s  F s 1
The controller is so simple and come to the form of PID controller:
0.25s 2  0.75s  0.5
c

(2   ) s( s  1)
2
By equating K  c , PID parameters is obtained from the above
IMC controller. As we see, all of the parameters are just depend
on  .
0.75 0.5 0.25 
kP  , kI  , kD  ,F 
2 2 2 2
32
Example 2
Step response for different values of  is plotted.
Both robustness and performance is achieved.

  10   0.33

 1  330.2
IMC based PID for different plants

 The IMC controllers shown in next slide Table, is designed via


the standard procedure developed in the previous example.
IMC leads to PID controllers for virtually all models common
in industrial practice. Note that the table includes systems with
pure integrators and RHP zeros. Occasionally, the PID
controllers are augmented by a first-order lag.

1 1 2 2
c  k c (1   D s  ) , 
 I s  F s 1   2

34
IMC based PID for different plants

35
IMC based PID for different plants

36
 IMC based PID
 Introduction
 General Relationships

 Example 2: PID Design

 Example 3: Model with high uncertainty


 Part 2: Continuous MIMO systems
 IMC Structure - MIMO Case
 Internal stability for stable plants
 General Internal stability

 IMC Design Procedure - MIMO Case


 Inner-Outer factorization

 Nominal Performance

 Robust Stability & Performance

 Exercises
37
Example 3: Model with high uncertainty
2e-2.45s
Acutal plant : p
0.5s  1
1
Plant model : p
s (s  5)

• Checking the uncertainty

Nyquist Diagram for p Nyquist Diagram for p 38


Example3

Step 1 (Nominal Performance):


q  (pMw m )1 p A1w m *
 s (s  5)

Step 2 (Robust Stability & Performance):


1 s (s  5)
f= , q = qf =
(s  1) 2
(s  1)2

q s (s  5)
c 
1  pq (s  1)2  1
39
Example 3
As you see in these figures, step response for different values of  were
plotted. Because of high uncertainty, good performance did not achieve.

  2.5   3.3

 5   10
40
 IMC based PID
 Introduction
 General Relationships

 Example 2: PID Design

 Example 3: Model with high uncertainty


 Part 2: Continuous MIMO systems
 IMC Structure - MIMO Case
 Internal stability for stable plants
 General Internal stability

 IMC Design Procedure - MIMO Case


 Inner-Outer factorization

 Nominal Performance

 Robust Stability & Performance

 Exercises
41
IMC Structure- MIMO Case
 In MIMO case, we deals with transfer function matrices.

d
r y
Controller Plant

Model
d

C  Q (I  PQ )1
Q  C (I  PC )1
42
IMC Structure- MIMO Case
 Sensitivity function and Complementary sensitivity function

y  Tr  Sd Q

S  (I  PQ )(I  (P  P )Q ) 1
T  PQ (I  (P  P )Q ) 1

 If the model is perfect ( P  P )

~ ~ ~ ~
S  1  PQ , T  PQ

43
Internal Stability for stable plants
 Theorem: Assume that the model is perfect ( P  P ),
Then the IMC system is Internally stable IFF both the plant P
and controller Q are stable.

44
General Internal Stability

 Remember: All elements of the below matrix have to be stable

 PC (I  PC )1 (I  PC ) 1 P 
 
 C (I  PC )1  C (I  PC )1 P 

 In IMC Structure for P  P Q
By substituting C  Q (I  PQ )1

We have
 PQ (I  PQ )P 
 
Q  QP 
45
General Internal Stability
 PQ (I  PQ )P 
 
Q  QP 

 This implies that Q has to be stable and that in the elements of


above matrix the factor Q and I-PQ have to cancel any unstable
poles of P. Thus both Q and I-PQsmust have RHP zeros at the
plant RHP poles. Special care has to be taken to cancel these
1
common RHP zeros when the controller C  Q (I  PQ ) is
constructed. Minimal or balanced realization software can be
used to accomplish that.

46
 IMC based PID
 Introduction
 General Relationships

 Example 2: PID Design

 Example 3: Model with high uncertainty


 Part 2: Continuous MIMO systems
 IMC Structure - MIMO Case
 Internal stability for stable plants
 General Internal stability

 IMC Design Procedure - MIMO Case


 Inner-Outer factorization

 Nominal Performance

 Robust Stability & Performance

 Exercises
47
Inner-Outer factorization
 Theorem: let G (s )  C (sI  A )1 B  D be a minimal realization of
the square transfer matrix G(s) , and let G(s) have no zeros on the
1
iw-axis including infinity. Then we have G ( s )  N (s ) M (s )

where N,M are stable and N (i  )H N (i  )  I , and


N (s )  (C  QF )(sI  (A  BR 1F )) 1 BR 1  Q
M (s )1  F (sI  A )1 B  R where F  Q T C  (BR 1 )T X

with X the stabilizing [i.e., it makes (A  BR 1F ) stable] real symmetric solution
of the following algebraic Riccati equation (ARE):

(A  BR 1Q T C )T X  X (A  BR 1Q T C )  X (BR 1 )(BR 1 )T X  0


So that :
1
PA  N and PM  M
48
IMC Design – MIMO case

 IMC design procedure consists of two steps

 Step1: Nominal Performance


Q is selected to yield a good performance for inputs, without regard for
constraint and model uncertainty.

 Step2: Robust Stability & Performance


The Q obtained in step 1 is detuned to satisfy the robustness requirements.
For that purpose, Q is augmented by a filter F of fixed structure.

Q  Q (Q , F ) i .e : Q  QF

49
Step1: Nominal Performance

o The plant P can be factored as P  PA PM


 PA is stable allpass portion
 PM is MP portion

The procedure for carrying out this factorization is:


“Inner – Outer Factorization”
Define the square matrix V(s){set of n inputs vi }
V ( s )  (v1 ( s ), v 2 ( s ),, v n ( s ))
v i ( s ) is a vector that describes expected direction and
frequency content of i th input.
Similarly : V  VM V A 50
Step1: Nominal Performance

 To achieve nominal performance, the controller is:


Q  PM1W 1 W PA1 V M
of the operand, all term s involving the poles of

*
V M1
SSS are om itted.

Where the operator 


. * denotes that after a partial fraction
expansion of the operand, all terms involving the poles of
PA1 are omitted.

51
Step2: Robust Stability & Performance
 The controller Q is to be detuned through a lowpass filter
F , to satisfy robust stability and performance. So the tuned
controller is:
Q  QF
 The filter F(s) is chosen to be a diagonal rational function that
F (s )  diag f 1 (s ) , , f n (s )
satisfy:

o The controller Q  QF must be proper


o Internal stability
o Asymptotic setpoint tracking/disturbance rejection.

52
Step2: Robust Stability & Performance
F (s )  diag f 1 (s ) , , f n (s )

 Experience has shown that the following structure is


reasonable:
av 1 1,l s v 1 1   a1,l s  a0,l
f l (s ) 
Where (s  1)v v l 1

 v is pole-zero excess.
 K is the number of open RHP poles of P
 m 0l is the largest multiplicity of such pole in any element of the lth row
of V.
 v l  m 0l  k

53
Step2: Robust Stability & Performance
av 1 1,l s v 1 1   a1,l s  a0,l
f l (s ) 
(s  1)v v l 1
 The numerator coefficients can be computed from solving a
system of v l linear equations with v l unknowns.

f l ( i )  1, i  0,1, ,k
 j
 d f (s )  0, j  1, , m 0l  1
 ds j l
 s  0

where :
 i (i  1,..., k ) are open RHP poles of P
0  0
54
 IMC based PID
 Introduction
 General Relationships

 Example 2: PID Design

 Example 3: Model with high uncertainty


 Part 2: Continuous MIMO systems
 IMC Structure - MIMO Case
 Internal stability for stable plants
 General Internal stability

 IMC Design Procedure - MIMO Case


 Inner-Outer factorization

 Nominal Performance

 Robust Stability & Performance

 Exercises
55
Exercise 1: Robust controller design
4 s
~
model : p 
2 e  s  3
s  1s  2
Input : Step
Design an IMC controller based on plant model.
Assume that the real plant is :
2e 4.6 s s  3
Plant :p 
s  1s  2
1.find filter parameter ( ) range that stabilize the system?
2. find  in order to achieve 10% over shoot. plot step response and find rise time
3. what is the relationship between filter pole locatoin,
performance and robustness? why?
56
Exercise 2: Poor modeling impact

2e 4.6 s s  3
Assume that the real plant is :p 
s  1s  2
Input : step
- 1.5
Assume that we have modeled this plant with : ~
p
(0.2s  1)(0.3s  1)
1. find filter parameter range that stabilize the system.
2. find  in order to achieve 10% over shoot.
plot step response and find rise time.
3. compare results with Exercise 1, propose better models and
compare the results, what are the advantages of better system identification?
57
Exercise 3: PID design

Design a PID controller through IMC :


2
Plant Model : ~ p
(0.4 s  1)(0.2 s  1)
Input : Step
1 - first find ~
q , q , c then find PID and lag compensator
parameters in terms of .
2 - Install the PID controller on nominal plant,
plot step response for   10
does any  makes the system unstable? why?

58
References
[1] C. E. Garcia and M. Morari, "Internal model control. A unifying review and some new results,"
Industrial & Engineering Chemistry Process Design and Development, vol. 21, pp. 308-323,
1982.
[2] C. E. Garcia and M. Morari, "Internal model control. 2. Design procedure for multivariable
systems," Industrial & Engineering Chemistry Process Design and Development, vol. 24, pp.
472-484, 1985.
[3] C. E. Garcia and M. Morari, "Internal model control. 3. Multivariable control law computation
and tuning guidelines," Industrial & Engineering Chemistry Process Design and Development,
vol. 24, pp. 484-494, 1985.
[4] M. Morari and E. Zafiriou, Robust Process Control. New Jersey: Prentice-Hall, Inc., 1989.
[5] A. Porwal and V. Vyas, "Internal model control (IMC) and IMC based PID controller,"
Bachelor of Technology, Department of Electronics & Communication Engineering, National
Institute of Technology, Rourkela, 2010.
[6] D. E. Rivera, Internal Model Control: A Comprehensive View. Tempe, Arizona: Arizona State
University, 1999.
[7] D. E. Rivera, et al., "Internal model control: PID controller design," Industrial & Engineering
Chemistry Process Design and Development, vol. 25, pp. 252-265, 1986.

59
Thank You For
Your Attention

60

S-ar putea să vă placă și