Analysis and Design of Linear Control Systems, 01st.

Basic Feedback Loop

d
C ( s)
1

Analysis and Design of Linear Control System Part2Spring, 2012 Instructor: Prof. Masayuki Fujita (S5-305)
Lecture Time and Place: Tuesday, 10:4512:15, S224 Schedule: 5th, 12th, 19th, June, 3rd, 10th, 24th, 31st, July No class: 26th, June TA: Tatsuya Ibuki, Yasuaki Wasa (S5-302)

F (s)

u
y

P( s)

F =1
Controller

Process
Fig. 11.1

P( s) : Process

r : Reference signal d : Load disturbance

n : Measurement noise

C ( s) : Feedback block ( F ( s ): Feedforward block) e : Output error u : Control variable

: Process output : Measured signal

Control System Transfer Functions

y = +n

Gang of Four
y

d
e

= P + n = P(u + d ) + n F = 1 = PCe + Pd + n Controller = PC ( r y ) + Pd + n

C (s)

u
y

P( s)

C ( s)
1

u y

P( s)

Process
Fig. 11.1

(1 + PC ) y = PCr + Pd + n PC P 1 d+ n y= r+ 1 + PC 1 + PC 1 + PC
1 y 1 = 1 u e 1 1 PC + PC PC + PC C + PC C + PC 1 + PC P 1 + PC P 1 + PC 1 1 + PC PC 1 + PC P 1 + PC 1 1 + PC PC 1 + PC C 1 + PC C 1 + PC 1 1 + PC r d n

Controller Gang of Four

Process
Fig. 11.1

Complementary 1 Sensitivity T = PC Sensitivity S = ( d ) 1 + PC Function ( r ) 1 + PC Function

( d )
(11.1)

PS =

P 1 + PC

Load Sensitivity Function

( n u )

CS =

C 1 + PC

Noise Sensitivity Function

(11.3)

General Representation*
d
n

General representation
y

[Ex. 11.1]

r F =1

C (s)
1

u
y

P( s)

u
Process
Fig. 11.1

y
Fig. 11.2

C ( s) 1

P( s)

Controller

C
0 0 0 0

Fig. 11.1

d ( w1 ) ( z1 )
u

( z 2 ) n( w2 )

P( s)

C (s)
y

1 P= P P

1 P 1 1 P

Process

P( s ) =
Controller

1 sa
k (s a) s

r y

G yr = ?

Stable ?

C=C
Fig. 11.2

e( z3 ) r ( w3 )

d w= n r

v z= e

C (s) =

Gyr (s) = T (s) =

k ,k >0 s+k

k >0

Analysis and Design of Linear Control Systems, 01st.

Gang of Four

Sensitivity Function

S ( s) =

Load Sensitivity Function

s s+k

Complementary Sensitivity Function k

Internal Stability

T ( s) =

s+k k (s a) s+k
T

e C ( s) u P( s)

s PS ( s ) = ( s + k )(s a )
S

Noise Sensitivity Function

can be unstable
T

CS ( s ) =

1
S =

Fig. 11.1

Internal Stability : All of the Gang of Four are stable.

( d )

S CS PS

1 1 + PC

Sensitivity T Function ( r )

PS

CS

a = 1

unstable a = 1
Step response

( d )

PS =

P Load Sensitivity CS 1 + PC Function ( nu )

Frequency response

Well-posed: 1 + P ( )C ( ) 0

Complementary Sensitivity Function C Noise Sensitivity = 1 + PC Function (11.3)

PC 1 + PC

Youla Parameterization (stable process) (12.2)

Youla Parameterization (unstable process)

C =

All stabilizing controllers P ( s ) : Stable process Q(s) : Stable transfer function (parameter)

Q 1 PQ

(12.8)

Q( s )

P( s )
P( s)

P(s) =

[Ex.]

A( s ) A( s ), B ( s ) : Stable B(s)
1 ( s 1)( s 2)

1
Fig.12.8 (a)

P( s) =

A( s ) =

1 ( s 1)( s 2) , B(s) = ( s + 1) 2 ( s + 1) 2

Coprime Factorization

Gang of Four
S = 1 PQ PS = P(1 PQ) CS = Q T = PQ

IMC : Internal Model Control All of these 4 transfer functions are stable. All stabilizing controllers are parameterized by Q

AF0 + BG0 = 1
[Ex.]
A( s ) =

F0 ( s ) , G0 ( s ) : Stable

1 ( s 1)( s 2) , B( s) = ( s + 1) 2 ( s + 1) 2

F0 ( s ) =

s+6 19s 11 , G0 ( s ) = s +1 s +1

Youla Parameterization (unstable process)

Analysis and Design of Linear Control Systems

Instructor: Prof. Masayuki Fujita Schedule: 5th, 12th, 19th, June, 3rd, 10th, 24th, 31st, July (No class) 26th, June Units: 2 units Grading: Some Basic Reports (given in each class) No Exam Textbook: K. J. Astrom and R. M. Murray, Feedback Systems, Princeton University Press (2008)
http://www.cds.caltech.edu/~murray/amwiki/Main_Page

C=

F0 + QB G0 QA

(12.9)

: All stabilizing Q(s) : Stable controllers

[Ex.] P( s ) =

1 ( s 1)( s 2)

C (s) =

Exercise: Internal Stability Gang of Four

19s 11 (Q = 0) s+6

S=

B(G0 QA) AF0 + BG0

T=

A( F0 + QB) AF0 + BG0

Chapter 10, 11, 12 Web: Information on the lecture will be continuously uploaded http://www.fl.ctrl.titech.ac.jp/course/SC/12SC/12SC.html

A(G0 QA) PS = AF0 + BG0

B( F0 + QB) CS = AF0 + BG0

Analysis and Design of Linear Control Systems, 01st.

5th June, 1st Lecture

11 Frequency Domain Design
11.1 Sensitivity Functions (12.2 Youla Parameterization ) Keyword : Basic Feedback Loop, Gang of Four Internal Stability, Youla Parameterization 11.3 Performance Specifications (12.3 Performance in the Presence of Uncertainty) (11.5 Fundamental Limitation) Keyword : Sensitivity Function

11.3 Performance Specifications Disturbance Attenuation ~ Open Loop vs. Closed Loop ~ Open Loop

Closed Loop (feedback)

dy G yd = P dy
G yd =

P 1 + PC = PS

The sensitivity function S ( j ) tells how the variations in the outputs are influenced by feedback

S ( j ) < 1

Disturbances are attenuated

Process Variations (12.3)

Process Variations (12.3)

C ( s) 1

dP

P( s)

C ( s) 1

dP

P( s)

Fig. 11.1

dy

ry
(12.11)

Fig. 11.1

P G yd = = PS 1 + PC

dG yd
Exercise

G yd

dP =S P

G yr =

PC =T 1 + PC

dG yr
Exercise

G yr

=S

dP P

(12.15)

The response to load disturbances is insensitive to process variations for frequencies where S ( j ) is small

Benefits of Feedback Disturbance Attenuation Insensitivity to Plant Variations Stabilization

Waterbed Effect (11.5) If the loop transfer function has no right half-plane poles

Bodes Integral Formula (11.5) Theorem 11.1 Bodes Integral Formula Assume that the loop transfer function L( s ) of a feedback system goes to zero faster than 1 / s as s , and let S ( s ) be the sensitivity function. If the loop transfer function has poles pk in the right half-plane, then the sensitivity function satisfies the following integral:

log S ( j ) d = 0 Conservation Law

log S < 0 S < 1

log S > 0 S > 1

There exists a range of freq. Waterbed Effect such that S > 1

log S ( j ) d =

log

1 d = p k 1 + L( j )
Im

(11.19)

Fundamental Limitation RHP poles fast (big): worse slow (small): better

Pole

fast

Re

slow

Analysis and Design of Linear Control Systems, 01st.

Benefits of Feedback Disturbance Attenuation Insensitivity to Plant Variations Sensitivity function S ( j )

Ms < 2

Sensitivity function S ( j )

S ( j )

1 S ( j ) < 1 S = 1 + L ( j ) > 1 1+ L

bs
A

[dB]

sm : the shortest distance

bs : Sensitivity Bandwidth Frequency 1 S ( j ) = (3 [dB])

from the Nyquist curve to the critical point (-1)

M s = 1 / sm : maximum sensitivity
(unstable)

[Ex.] 2 M s: Maximum Peak Magnitude of S ( j ) M s = max S ( j ) M s < 2 A : Maximum Steady-state Tracking Error Type 1: A = 0

sm : stability margin

Ms < 2
0.5 < sm < 0.8
Fig. 11.6(b)

5th June, 1st Lecture

11 Frequency Domain Design
11.1 Sensitivity Functions (12.2 Youla Parameterization ) Keyword : Basic Feedback Loop, Gang of Four Internal Stability, Youla Parameterization 11.3 Performance Specifications (12.3 Performance in the Presence of Uncertainty) (11.5 Fundamental Limitation) Keyword : Sensitivity Function

12th June, 2nd Lecture

12 Robust Performance
12.1 Modeling Uncertainty (12.3 Performance in the Presence of Uncertainty) (11.1 Sensitivity Function) (11.5 Fundamental Limitation) Keyword : Complementary Sensitivity Function Modeling Uncertainty 12.2 Stability in the Presence of Uncertainty (9.3 Stability Margins) (9.5 Generalized Notions of Gain and Phase) Keyword : Robust Stability