Lecture 4 - Simple Design Problems 1.

Introduction

K. J. Åström • The role of abstractions

Solve an abstract problem and you have the solution to
1. Introduction many concrete problems
2. PI Control of First Order Systems But you must be able to translate!
3. Properties of Second Order Systems Technology transfer
4. PD Control of Second Order Systems • Standard models e.g. in the form of transfer functions
5. PID Control of Second Order Systems b bs + b2
G (s) = , G ( s) = 2
6. Matlab s+a s + a1 s + a 2
• Essential that storage of mass, momentum and energy
7. Summary
can be captured by a few parameter
Theme: Solve simple design problems in a unified way. • Description of storage more complicated if variations are
Dynamics of simple closed loop systems. Matlab. rapid
– Moderate requirements of control performance

Systems of First Order PI Control of First Order System

• Many systems are approximately of first order • Transform the physical problem to a standard model e.g.
• The key is that the storage of mass, momentum and b
energy can be captured by one parameter P( s ) =
s+a
• Examples
• Pick a controller e.g. PI
– Velocity of car on the road
– Control of velocity of rotating system ki
C ( s) = k +
s
– Electric systems where energy storage is essentially in
one capacitor or on inductor
• Design a controller e.g. by pole placement
– Incompressible fluid flow in a pipe
• Closed loop system is of second order
– Level control of a tank
– Pressure control in gas tank • Translate results back to the physical system
– Temperature in a body with essentially uniform temper-
ature distribution (e.g. steam filled vessel)

c K. J. Åström September, 2000 1
 PI Control of First Order System Behavior of Second Order System
Process The system

 ω0 1 − ζ 2
b ω 20 ω0
P( s) = G ( s) =
s+a s2 + 2ζ ω 0 s + ω 20
−ζ ω 0
Controller has the step response ζ = 0 (dashed), 0.1,
ki 0.2, 0.5, 0.707, 1 (dashed-dotted) 
C( s) = k +
s 2

Transfer functions 1.5

Y ( s) PC b( ks + ki )
= = 2 1
R( s) 1 + PC s + ( a + bk) s + bki
0.5

The closed loop can be given arbitrary values. The values

k = (2ζ ω 0 − a)/b and k i = ω 02/b give the characteristic
0

ω 0t
0 5 10 15

polynomial s2 + 2ζ ω 0 s + ω 02.

Generating Step Responses Behavior of Second Order System with a Zero

t=0:0.1:6.5*pi;

 ω0 1 − ζ 2
z=0; The system ω0
sys=tf(1,[1 2*z 1]);
y0=step(sys,t); ω 0 ( s + αω 0 ) ○
G ( s) =
z=0.1; s2 + 2ζ ω 0 s + ω 20 −ζ ω 0
sys=tf(1,[1 2*z 1]);
y1=step(sys,t); 
has the step response α = 0.2, 0.5, 1 (dashed), 2, 5 and ∞.
...
3
plot(t,y0,’r--’,t,y1,’b-’)
2.5
axis([0 15 0 2])
2
xlabel(’x’)
1.5
ylabel(’y’)
1

0.5

ω 0t
0 1 2 3 4 5 6 7 8 9 10


c K. J. Åström September, 2000 2
 Behavior of Second Order System with a Zero Systems of Second Order
• Storage of mass, momentum and energy can be captured
ω0 s + αω 0
G ( s) = by two parameter
α s2 + 2ζ ω 0 s + ω 20
• Examples
α = 0.2, 0.5 1 (dashed), 2, 5 and ∞.
– Position of car on the road
3
– Control of angle of rotating system
2.5

2
– Stabilization of satellites
1.5 – Electric systems where energy is stored in two ele-
1 ments (inductors or capacitors)
0.5
– Levels in two connected tanks
0

ω 0t – Pressure in two connected vessels

0 1 2 3 4 5 6 7 8 9 10

Why do all curves intersect at the same point?

The Audience is Thinking ...

PI Control of Second Order System Behavior of Third Order System

Process The system
b1 s + b2 αω 30
P( s ) = G ( s) =
s2 + a1 s + a2 ( s2 + 2ζ ω 0 s + ω 20 )( s + αω 0 )
Controller has the step response α = 0.2, 0.5, 1 (dashed), 2 and ∞.
ki
C( s) = k + 1.2
s
1

Closed loop transfer function from reference r to output y 0.8

0.6
Y ( s) PC ( ks + ki )( b1 s + b2 )
= = 3 0.4
R( s) 1 + PC s + ( a1 + b1 k) s2 + ( a2 + b1 ki + b2 k) s + b1 ki
0.2

The poles cannot be chosen arbitrarily. A more complex 0

0 1 2 3 4 5 6 7 8 9 10
controller is required to have full freedom ω 0t


c K. J. Åström September, 2000 3
 PD Control of Second Order System PD Control of Second Order System ...
Process The closed loop transfer function
b Y ( s) PC b( kd s + k) b( kd s + k)
P( s ) = = = 2 = 2
s2 + a1 s + a2 R( s ) 1 + PC s + a1 s + a2 + b( kd s + k) s + ( a1 + bkd ) s + a2 + bk

Notice that polynomial B is of zero degree. Controller

The zero s = − k/ k d may cause a considerable overshoot if it is
C( s) = k + kd s
closer to the origin than the dominant poles.
Closed loop transfer function from reference to output This zero is caused by the controller. Can the controller be
modified so that the zero is changed or removed?
Y ( s) PC b( kd s + k) b( kd s + k)
= = 2 = 2
R( s) 1 + PC s + a1 s + a2 + b( kd s + k) s + ( a1 + bkd ) s + a2 + bk The Audience is Thinking ...

Complexity of controller sufficient to place all the poles.

Systems with Two Degrees of Freedom PID Control of Second Order System
Changing the controller to b
Process: P(s) =
s2 + a1 s + a 2
U ( s) = kE( s) − kd sY ( s) = kR( s) − ( k + kd s) Y ( s)
ki
Controller: C(s) = k + + kds
The signal paths from reference R to control U is different s
from signal path from ouput Y to control. Derivative acts only Closed loop transfer function from reference r to output y
on the output.
Y ( s) PC b( kd s2 + ks + ki )
Y ( s) = =
= 2
k R( s) 1 + PC s( s2 + a1 s + a2 ) + b( kd s2 + ks + ki )
R( s ) s + ( a1 + kd ) s + a2 + k b( kd s2 + ks + ki )
=
s3 + ( a1 + bkd ) s2 + ( a2 + bk) s + bki

All poles can be chosen arbitrarily. Since the zeros comes from
the controller they can also be chosen arbitrarily by choosing a
structure with two degrees of freedom!


c K. J. Åström September, 2000 4
 Computational Tools Summary
Analysis and design of control system has been drastically • Linearize the process model
influenced by good computational tools. • Select a controller of sufficient flexibility (complexity)
• Matlab and its development • Derive the closed loop characteristic equation
– A matrix calculator • Pick controller parameters to give desired characteristic
– APL a precursor equations (pole placement). Typical prototypes:
– Cleve Moler s2 + 2ζ ω 0 s + ω 02 = 0
– CTRL-C, MatrixX, PC Matlab (s2 + 2ζ ω 0 s + ω 02)(s + ω 0 ) = 0
• Public domain versions: Octave and SciLab • Insight into behavior of low order systems
• Interactive version: SysQuake • A little insight about effect of closed loop zeros
• Many simple problems can be solved in this way
• BUT are the solutions good?


c K. J. Åström September, 2000 5