Lecture 18
Practical Issues in
PID Implementation
CL-417 Process Control
Prof. Kannan M. Moudgalya
IIT Bombay
Tuesday, 4 September 2012
1/25 CL-417 Process Control Practical Issues in PID Implementation
Outline
Implementable derivative mode
Discretisation (minimal)
Handling communication mismatch
2/25 CL-417 Process Control Practical Issues in PID Implementation
PID controller
PID controller has proportional, integral and
derivative modes:
u(t) = u + K
c
_
e(t) +
1
i
_
t
0
e(w)dw +
d
de(t)
dt
_
e is error, u is control eort and u is the
steady state (or bias) value of u
It has three tuning parameters, K
c
,
i
,
d
Recall the tuning guidelines to change these
parameters
3/25 CL-417 Process Control Practical Issues in PID Implementation
Summary of PID Controller
Proportional mode:
Most popular control mode. Increase in
proportional mode generally results in
Decreased steady state oset
Increased oscillations
Integral Mode:
Used to remove steady state oset. Increase
in integral mode generally results in
Zero steady state oset
Increased oscillations
4/25 CL-417 Process Control Practical Issues in PID Implementation
Summary of PID Controller
Derivative Mode:
Mainly used for prediction purposes
Increase in derivative mode results in
Decreased oscillations and improved stability
Sensitive to noise
Used extensively by EE/ME
Less used by chemical engineers, mainly
because of noise problems
PID is most popular in industry
5/25 CL-417 Process Control Practical Issues in PID Implementation
How does one design the PID controller?
By direct synthesis - will study later in this
course
Using advanced control techniques and
implementing them as PID
By trial and error, using tuning methods
But need a starting point
Will study the most popular tuning method
6/25 CL-417 Process Control Practical Issues in PID Implementation
Reaction Curve - Ziegler Nichols Tuning
Applicable only to stable systems
Give a unit step input and get
R = K/
L
K
1. the time lag after which the system starts
responding (L),
2. the steady state gain (K) and
3. the time the output takes to reach the steady
state, after it starts responding ()
7/25 CL-417 Process Control Practical Issues in PID Implementation
Reaction Curve - Ziegler Nichols Tuning
R = K/
L
K
Calculate slope, R = K/
8/25 CL-417 Process Control Practical Issues in PID Implementation
Reaction Curve Method Ctd.
R = K/
L
K
PID settings are given below:
K
c
i
d
P 1/RL
PI 0.9/RL 3L
PID 1.2/RL 2L 0.5L
Consistent units should be used
9/25 CL-417 Process Control Practical Issues in PID Implementation
Stability Method - Ziegler Nichols Tuning
Another method to tune PID parameters:
Close loop with a proportional controller
Gain of controller is increased until the
closed loop system becomes unstable
At the verge of instability, note down
controller gain K
u
, period of oscillation P
u
PID settings are given below:
K
c
i
d
P 0.5K
u
PI 0.45K
u
P
u
/1.2
PID 0.6K
u
P
u
/2 P
u
/8
Consistent units should be used
10/25 CL-417 Process Control Practical Issues in PID Implementation
Practial Issues: 1. Noise
11/25 CL-417 Process Control Practical Issues in PID Implementation
Handling noise in error signal
PID control law is given by
u(t) = K
c
_
e(t) +
1
i
_
t
0
e(w)dw +
d
de(t)
dt
_
Problem with derivative, thanks to noise
Can be seen in Laplace transformed version
also:
U(s) = K
c
(1 +
1
i
s
+
d
s)E(s)
Can see that the derivative mode results in a
high pass transfer function.
12/25 CL-417 Process Control Practical Issues in PID Implementation
Implementable PID Controller
Recall U(s) = K
c
_
1 +
1
i
s
+
d
s
_
E(s)
Problem happens at high s in the derivative
mode: magnitude is unbounded
Constrain its magnitude at high frequency
Filtered derivative mode:
u(t) = K
c
_
1 +
1
i
s
+
d
s
1 +
d
s
N
_
e(t)
N is a large number
Typical values are 5, 10, etc.
13/25 CL-417 Process Control Practical Issues in PID Implementation
PID Control Implementation
Discrete version is required for
implementation
Many ways to do this
Digital Control by Moudgalya presents
several methods
We present one approach now
14/25 CL-417 Process Control Practical Issues in PID Implementation
Practial Issues: 2. Discretisation
15/25 CL-417 Process Control Practical Issues in PID Implementation
PID Controller - Discretisation
Recall the PID control law:
u(t) = K
c
_
e(t) +
1
i
_
t
0
e(t)dt +
d
de(t)
dt
_
Let us discretise, with u(n) = u(t)|
t=n
u(n) = K
c
[e(n)+
1
i
_
T
s
e(0) + e(1)
2
+ + T
s
e(n 1) + e(n)
2
_
+
D
e(n) e(n 1)
T
s
_
16/25 CL-417 Process Control Practical Issues in PID Implementation
Practial Issues: 3. Communication errors
Example:
Consider manual to automatic mode change
Suppose that the controller thinks that the
valve position oset is 0%
Suppose that the control law is calculated as
+0.1%
But the actual control valve position is at
+10% oset and that it is working ne
With the new control eort, the valve will
swing from +10% oset to +0.1% oset
How do we minimise this problem, given that
there is a communication mismatch?
17/25 CL-417 Process Control Practical Issues in PID Implementation
Calculate Change in PID Control
u(n) = K
c
[e(n)+
1
i
_
T
s
e(0) + e(1)
2
+ + T
s
e(n 1) + e(n)
2
_
+
D
e(n) e(n 1)
T
s
_
Calculate u(n) u(n 1) and simplify
u(n) u(n 1) = s
0
e(n) + s
1
e(n 1) + s
2
e(n 2)
s
0
= K
c
_
1 +
T
s
2
i
+
d
T
s
_
, s
2
= K
c
d
T
s
s
1
= K
c
_
1 +
T
s
2
i
2
d
T
s
_
18/25 CL-417 Process Control Practical Issues in PID Implementation
Bumpless Control
From the previous slide
u(n) u(n 1) =
s
0
e(n) + s
1
e(n 1) + s
2
e(n 2)
Because control law is given in terms of u,
Using absolute values could give a bump
In the previous example, the control eort
will go from 10% to 10.1%
Mismatch between the state of end control
element and what controller thinks it to be,
may not matter much
Hence known as bumpless transfer
19/25 CL-417 Process Control Practical Issues in PID Implementation
Practical Issues: 4. Setpoint Kick
To be covered later in this course
20/25 CL-417 Process Control Practical Issues in PID Implementation
Problem 3 I
Corridor Consultants Ltd. has asked two of its
engineers Ram and Mala to design a PD
controller for the plant G(s)
G(s) =
1
(s + 1)(s + 3)
with the following requirements: (1) the closed
loop system should respond reasonably fast (2)
the closed loop system should be stable for large
values of the proportional controller gain as well.
21/25 CL-417 Process Control Practical Issues in PID Implementation
Problem 3 II
Ram designs the following controller
G
cr
= K
c
(1 + 0.2s)
and Mala designs the following controller
G
cm
= K
c
(1 + 0.5s)
1. Which of the two controllers is better and
why?
22/25 CL-417 Process Control Practical Issues in PID Implementation
Problem 3 III
2. At the time of implementation of the
controller, however, it is found that the
actual model of the plant, which we denote
as G
(s), is given as
G
(s) =
1
s(s + 1)(s + 3)
Which of the two controllers would you
prefer now? Why?
Hint: Qualitative arguments, possibly with root
locus with a few calculations, could be sucient
most of the time. (40)
23/25 CL-417 Process Control Practical Issues in PID Implementation
What we learnt today
Implementable derivative mode
Discretisation
Handling communication mismatch
Root locus problem from a previous exam
24/25 CL-417 Process Control Practical Issues in PID Implementation
Thank you
25/25 CL-417 Process Control Practical Issues in PID Implementation
