You are on page 1of 48

MCH 3008 CONTROL SYSTEMS

LECTURE 9: PID CONTROL


Dr. Beste Bahçeci
Bahçeşehir University
Overview of Lecture 7
• PID control
• Ziegler Nichols Method
PID Control PID Dc ( s ) 
U (s) k
 kP  kD s  I
Proportional E ( s) s
Derivative P U ( s)
Dc ( s )   kP
Integral E ( s)
U ( s)
PD Dc ( s )   kP  kD s
E ( s)
U (s) kI
PI Dc ( s )   kP 
E ( s) s

e u Plant
Dc (s ) G
A
Proportional Control (P Control) G( s)  2
s  a1s  a2
U ( s)
Dc ( s )   kP u (t )  k P e(t )
E ( s)
A
kP
Y ( s) Dc ( s )G ( s )  s 2  a1s  a2 kP A
 
R( s ) 1  Dc ( s )G ( s ) 1  k A s 2  a1s  a2  k P A
P 2
s  a1s  a2
L 
This is a type 0 system. 2 n  2
n

Proportional control increases rise speed, but can cause low damping.
e u Plant
Dc (s ) G
Remember the DC motor dynamics. T  K t ia

dia
La  Ra ia  va  K em J mm  bm  K t ia
dt
It has a second order characteristic equation
between the input voltage and the motor
velocity:
 m ( s) Kt

Va ( s ) s( J m s  b)( La s  Ra )  K t K e 

 m ( s) Kt


Va ( s ) s J m La s 2  ( J m Ra  bLa ) s  bRa  K t K e 
Kt Voltage – position
 m ( s) J m La relation

Va ( s )  2 ( J m Ra  bLa ) bRa  K t K e 
s s  s
 J m La J m La 
Kt
 m ( s) J m La
 Voltage – velocity
Va ( s )  2 ( J m Ra  bLa ) bRa  K t K e 
 s  s   relation
 J m L a J m L a 
Kt
 m ( s) J m La
 Voltage – velocity
Va ( s )  2 ( J m Ra  bLa ) bRa  K t K e 
 s  s   relation
 J m La J m La 
Kt
 m ( s) J m La

Va ( s )  2 ( J m Ra  bLa ) bRa  K t K e 
 s  s  
 J L
m a J L
m a 

A
Note that it is in this form: G ( s ) 
s 2  a1s  a2
We will use this model in P control simulations.

e u Plant
Dc ( s )  k P G
 m (s) A
G( s)   2
Va ( s ) s  a1s  a2
s 2

 a1s  a2  m ( s )  AVa ( s )
m (t )  a1 m (t )  a2m (t )  AVa (t )
m (t )  AVa (t )  a1 m (t )  a2m (t )

Va (t ) m (t )  m (t ) m (t )
1 1
A
s s


a1

a2
e u Plant
Dc ( s )  k P G

e u Plant
Dc ( s )  k P G
e u Plant
Dc ( s )  k P G
Reference

Actual
kP  3
Higher gains
reduce
steady state
error, and
Y ( s) kP A speed up the
 2 response,
R( s ) s  a1s  a2  k P A
but they end
2 n k P  10 up in larger
n2
k P  3 oveshoots
too. There is
kP  1
always a
steady-state
error. This is
something
we would
expect from
a type 0
system.
Proportional-Derivative Control (PD Control) A
G( s) 
U ( s) s 2  a1s  a2
Dc ( s )   kP  kD s u (t )  k P e(t )  k D e(t )
E ( s)
A
Y ( s) Dc ( s )G ( s )
k P  k D s 
 s 2
 a1s  a2

R( s ) 1  Dc ( s )G ( s ) A
1  k P  k D s  2
s  a1s  a2
 2
k P  k D s A
s  a1  k D As  a2  k P A
With PD control speed and damping of the response can be adjusted.
e u Plant
Dc (s ) G
Ther are two types of PD control:

PD control with position feedback

PD control with rate (and position) feedback


kP  3
k D  0.0003
kP  3
kD  0

Derivative control action


adds damping to the
kP  3 system without
k D  0.0003 improving the steady
state behavior.
Proportional-Integral Control (PI Control) G( s) 
A
U (s) k t s 2  a1s  a2
Dc ( s )   kP  I u (t )  k P e(t )  k I  e( )d
E ( s) s 0

 kI  A
 P
k   2
Y ( s) Dc ( s )G ( s )  s  s  a1s  a2
 
R( s ) 1  Dc ( s )G ( s )  k  A
1  kP  I  2
 s  s  a1s  a2
Open loop
 kP s  kI  A transfer
  2
 s  s  a1s  a2 function.

 k s  kI  A
1  P  2
 s  s  a1s  a2
With PI control the system type can be changed.
In this example, the system type is changed form 0 to 1.
Zero steady state error with a step reference is expected.
k P  10
k I  5000
k P  10
k I  5000

k P  10
The integral action kI  0
removes the steady
state error. It also
increases overshoot
slightly.
Proportional-Derivative-Integral Control (PID Control) A
G( s) 
s 2  a1s  a2
U (s) k t
Dc ( s )   kP  kD s  I u (t )  k P e(t )  k D e(t )  k I  e( )d
E ( s) s 0

 kI  A
 kP  kD s   2
Y ( s) Dc ( s )G ( s )  s  s  a1s  a2
 
R( s ) 1  Dc ( s )G ( s )  k  A
1   kP  kD s  I  2
 s  s  a1s  a2
Open loop
 kD s 2  kP s  kI  A
  2 transfer
 s  s  a1s  a2 function.

 kD s 2  kP s  kI  A
1    2
 s  s  a1s  a2

With PID control the system type can be changed too.


Y ( s)

R( s)

 kD s 2  kP s  kI  A
  2
 s  s  a1s  a2

 kD s 2  kP s  kI  A
1    2
 s  s  a1s  a2
 kD s 2  kP s  kI  A
  2
Y ( s)
 
s  s  a1s  a2
R( s)  kD s 2  kP s  kI  A
1    2
 s  s  a1s  a2

k D As 2  k P As  k I A
 3
s  a1  k D As 2  a2  k p As  k I A

We can adjust three coefficients.


 All the roots of the third order polynomial can
be selected by the control designer.
k P  10
k I  5000
k D  0.003
Good damping and zero
steady state error can be
achieved with PID control.
k P  10
k I  5000
kD  0 k P  10
k I  5000
k D  0.003
Remember feedback control system with unity feedback loop equations:

Ycl

DG G DG D DG D
Ycl  R W V U R W V
1  DG 1  DG 1  DG 1  DG 1  DG 1  DG
1 G DG
Ecl  R  Ycl  R W V
1  DG 1  DG 1  DG
Remember that the system type can be defined with respect to reference
and disturbance signals. Integral action, bu changing the disturbance-to-
error system type can improve the disturbance rejection too.
Proportional-Derivative-Integral Control (PID Control) and disturbance
rejection
A
G( s)  2
s  a1s  a2
U ( s) k
Dc ( s )   kP  kD s  I
E ( s) s Step
t
w(t )
u (t )  k P e(t )  k D e(t )  k I  e( )d
0

Dc (s )
0
Dc (s )
0
k P  10
k I  5000
k D  0.003

Step disturbance rejected by PID controller.


k P  10
kI  0
Step disturbance is not rejected k D  0.003
when the integral action is cancelled.
Ziegler-Nichols tuning of PID controllers
There many detailed design procedures for plants with detailed models. If
you do not have such a model then experimental tuning techniques can be
applied.

Ziegler-Nichols techniques are experimental techniques for PID controller


tuning.

There are two Ziegler-Nichols techniques:

1) Based on a “first order system with delay” assumption.


2) Based on ultimate sensitivity.
Ziegler-Nichols technique based on a “first order system with delay”
assumption Y ( s ) Ae  std

U ( s ) s  1

Step response experiment


is caried out firstly.
Slope R and delay (lag) L are
measured from experimental data.

The PID controller parameters are


obtained as functions of R and L.
The PID gain functions are obtained for
a quarter decay ratio.

The decay ratio is the


amplitude ratio measured in
peaks which follow earch
other.

A signal with quarter decay ratio is


shown in this figure.
Remember our PID controller formula:
U ( s) kI
Dc ( s )   kP  kD s 
E ( s) s

Ziegler-Nichols rules use an alternative PID controller formula:

U ( s)  1 
Dc ( s )   k P 1  TD s  
E ( s)  TI s 

Proportional gain
Reset rate
Derivative rate
U ( s)  1 
Ziegler - Nichols rules for the regulator Dc ( s)   k P 1  TD s  ,
E (s)  TI s 
for a decay ratio of 0.025 :

Type of controller Gain


P k P  1 RL
PI k P  0.9 RL
TI  L 0.3
PID k P  1.2 RL
TI  2 L
TD  0.5 L
Example: Ziegler-Nichols quarter decay ratio tuning
Obtain P and PI controllers.

Plant: Heat exchanger

Tm ( s ) Ke  td s

As ( s ) Rms cvs mwcvw s 2  ( Rms cvs ww cvw  ms cvs  mwcvw ) s  wwcvw
Step response result:

1
R
90

Type of controller Gain


P k P  1 RL
PI k P  0.9 RL
L  13s TI  L 0.3
P k P  1 RL  6.92
PI k P  0.9 RL  6.22
TI  L 0.3  43.3

Resulting step responses Resulting step responses after


further manual tuning
Ziegler-Nichols technique based on ultimate sensitivity
Experiment:
Connect a unity feedback loop. Employ proportional control.
Increase the proportional gain until sustained oscillations are
observed at the output.
Record proportional gain. Call it K u .
Record the period of oscillations at the output. Call this period Pu .

PID controller parameters are obtained as functions of K u and Pu .

Pu
U ( s)  1 
Ziegler - Nichols rules for the regulator Dc ( s)   k P 1  TD s  ,
E (s)  TI s 
with ultimate sensitivity rules :

Type of controller Gain


P k P  0 .5 K u
PI k P  0.45 K u
TI  Pu 1.2
PID k P  0 .6 K u
TI  0.5 Pu
TD  0.125 Pu
Example: Ziegler-Nichols quarter ultimate sensitivity tuning
Obtain P and PI controllers.

Plant: Heat exchanger

Tm ( s ) Ke  td s

As ( s ) Rms cvs mwcvw s 2  ( Rms cvs ww cvw  ms cvs  mwcvw ) s  wwcvw
K u  15.3
Pu  42 s
Resulting step responses Resulting step responses after
further manual tuning
Remark: P, PI, PD, PID controllers are commnly used. I controllers are
seldom used. D controllers are not used because they amplify sensor
noise.
Integrator overload
PI and PID controllers can have problems because of actuator saturation.

Suppose that the actuator is saturated and error is nonzero. Then the
error integral (and hence the integral action in the control signal) will be
unbounded.

The solution is “integrator anti-windup”:


Stop integration when the controller operates at its saturation limit.

You might also like