Gain Scheduling

Bo Bernhardsson and Karl Johan strm

Department of Automatic Control LTH,

Lund University

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Gain Scheduling

What is gain scheduling ?

How to find schedules ?
Applications
What can go wrong ?
Some theoretical results
LPV design via LMIs
Conclusions

To read:
Leith & Leithead, Survey of Gain-Scheduling Analysis & Design
To try out:
Matlab - Gain Scheduling

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Gain Scheduling
Controller
parameters Gain
schedule

Operating
condition
Command
signal Control
signal
Controller Process Output

Example of scheduling variables

Production rate
Machine speed, e.g. DVD player
Mach number and dynamic pressure

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 How to Find Schedules?

Select scheduling variables: Variable(s) should reflect changes in

system dynamics.
Make (linear) control design for different operating conditions:
For instance with automatic tuning
Use closest control design, or interpolate: Many ad-hoc or
theoretically motivated methods exist
Verify performance: Simulations. Some methods exist that
guarantee performance; usually conservative though

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Scheduling on controller output

GS Ref

FIC

FT

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Nonlinear Valve
A typical process control loop

Process

uc c u v y
PI f 1 f G 0 (s)

Valve characteristics and a crude approximation

10 f(u)
f (u)
0
0 0.5 1 1.5 2
Time

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Results

Without gain scheduling

Loop is either too slow or unstable

With gain scheduling

uc
0.3

0.2
y
0 20 40 60 80 100

uc Time

1.1

1.0
y
0 20 40 60 80 100
Time
uc
5.1

5.0 y
0 20 40 60 80 100
Time

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Schedule on Process Variable

GS Ref

LIC

LT

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Schedule on External Variable

GS Ref

TIC

TT FT

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Concentration Control

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Concentration Control

Performance with changing flow Vd

(a) Output c
cr
1.0
q = 0.5
0.5 q = 0.9
q = 1.1
0.0
q =2
0 5 10 15 20
Time
(b) Control signal cin
1.5 q = 0.9q = 0.5
1.0

0.5 q = 1.1
q=2
0.0
0 5 10 15 20
Time

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Variable Sampling Rate

Process model
1 Vm Vd
G(s) = es , T = ,
1 + sT q q

Sample system with period

Vd
h=
nq
Sampled model becomes linear time-invariant

c(kh+h) = ac(kh)+(1a)u(khnh), a = eqh/Vm = eVd /(nVm )

Sampled equation does not depend on q !!

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Results

Digital control with h = 1/(2q).

The flows are: (a) q = 0.5; (b) q = 1; (c) q = 2

(a) 1 1

c cin
0 0
0 5 10 15 20 0 5 10 15 20
Time Time
(b)
1 1
cin
c
0 0
0 5 10 15 20 0 5 10 15 20
Time Time
(c)
1 1

c cin
0 0
0 5 10 15 20 0 5 10 15 20
Time Time

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Flight control

Pitch dynamics
q =

Nz e

Operating conditions
80

60
Altitude (x1000 ft)

40
3 4

20

1 2
0
0 0.4 0.8 1.2 1.6 2.0 2.4
Mach number

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 The Pitch Control Channel
VIAS H

Pitch stick
Gear
VIAS
Position
Filter A/D K DSE

H M
H

T1s
K SG
1+ T1s

D/A
Acceleration
Filter A/D D/A
To servos
M Filter
Pitch rate

1
Filter A/D K Q1 K NZ
1+ T3 s

M H
T2 s
K QD
1+ T2 s

H M VIAS

Many scheduling variables

Bo Bernhardsson and Karl Johan strm Gain Scheduling

Schedule of KQ wrt airspeed (IAS) and height (H)

K QD IAS K QD H

0.5
0.5

0 0
0 1000 0 10 20
V IAS (km/h) H (km)

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Surge Tank Control

A surge tank is used to smooth flow variations. The is allowed will

fluctuate substantially but it is important that the tank does not become
empty or that it overflows.

LT LIC

Surge tank

FIC

FT

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 The Igelsta Power Station

Controller structure before modification

PIC

PT TIC

TT
Set point
Fuel/air

DH
network

Boiler Heat exchanger

Modified controller structure

PIC

PT TIC

TT XT
Set point
Fuel/air

DH
network

Boiler Heat exchanger

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 What can go wrong?

Most designs are done for the time-frozen system, i.e. as if scheduling
parameter is constant.
Theory and practice: This will work also when (t) is slowly varying.
But can go wrong for fast varying parameters.
Following example is from Shamma and Athans, ACC 1991

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Shamma - Athans

Resonant system with varying resonance frequency

1 1 1
G (s) = 2 = C(sI A())1 B
s s + 0.2s + 1 + 0.5(t) s
with 1 (t) 1
Controller design: LQG + integrator on system input, LQG parameters

Q11 = C T C, Q22 = 108 , R11 = B2 ()B2 ()T , R22 = 102

Gives controller K (s) with good robustness margins when constant

Frozen-system loop-gain G (s)K (s) is actually independent of
(using B2 ()T = [1 0 1 + 0.5])

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Shamma-Athans

Step response and GOF for any constant look fine.

All give the same curves

1.2
2 2
10 10

PS
S
0 0
10 10

0.8
-2 -2
10 0 2
10 0 2
10 10 10 10

0.6

2 2
10 10
0.4

CS

T
0 0
10 10
0.2

-2 -2
10 0 2
10 0 2
0 10 10 10 10
0 0.5 1 1.5 2 2.5 3 w w

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Shamma-Athans
But if (t) = cos (2t) the system becomes unstable
2.5

1.5

0.5

-0.5
0 10 20 30 40 50 60 70 80 90 100

It can be shown that the open loop LTV system is unstable in this case.
Several theorems show that if varies slowly performance for the
frozen-system analysis is maintained for the true system

Small-gain theorem
Lyapunov theory using V = xT P x or V = xT P ()x
Bo Bernhardsson and Karl Johan strm Gain Scheduling
 Gain-scheduling design

Several authors have worked with systematic design methods

Shamma-Athans
Packard
Apkarian-Gahinet
Helmersson
...

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Gain-scheduling for LPV systems by LMIs

Apkarian, Gahinet (1995) A Convex Characterization of

Gain-Scheduled H Controller
Model assumption

x = A((t))x(t) + B((t))u(t)
y = C()x(t) + D((t))u(t)

Controller structure

(t) = AK ((t))(t) + BK ((t))y(t)

u(t) = CK ((t))(t) + DK ((t))y(t)

Will assume both process and controller depends on via a fractional

transformation.

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Main Idea

 T  T
q q q y w = Pa (s) w w w u u
Bo Bernhardsson and Karl Johan strm Gain Scheduling
 Main Result - a sufficency condition

The closed loop system is stable for all (t) with k(t)k < 1/ and
the L2 induced norm from w to q satisfies

max kTqw k <

k(t)k<1/ 2

if there is a scaling matrix L commuting with so that

! !
L1/2 0 L1/2 0
k Fl (Pa , K) k<
0 I 0 I

Sufficient, not necessary

The condition can be checked by an LMI, also gives the controller.

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 If Time Permits

http://se.mathworks.com/help/control/ug/gain-scheduled-control-of-a-
chemical-reactor.html

Bo Bernhardsson and Karl Johan strm Gain Scheduling

 Summary - Gain Scheduling

Very useful technique

Linearization of nonlinear actuators

Surge tank control
Control over wide operating ranges

Requires good models

Issues to consider

Choice of scheduling variable(s)

Granularity of tables, interpolation
Bumpless parameter changes

Bo Bernhardsson and Karl Johan strm Gain Scheduling