ELEC ENG 4CL4:
Control System Design
Notes for Lecture #33
Friday, April 2, 2004
Dr. Ian C. Bruce
Room: CRL-229
Phone ext.: 26984
Email: ibruce@mail.ece.mcmaster.ca
Chapter 10
Goodwin,
Graebe, Salgado, Prentice Hall 2000
3. Cascade Control
Next we turn to an alternative architecture for
dealing with disturbances. The core idea is to
feedback intermediate variables that lie between the
disturbance injection point and the output. This
gives rise to so called, cascade control.
Chapter 10
Goodwin,
Graebe, Salgado, Prentice Hall 2000
Cascade control is very commonly used in practice.
For example, if one has a valve in a control loop,
then it is usually a good idea to place a cascade
controller around the valve. This requires
measurements to be made of the flow out of the
valve (see next slide) but can significantly improve
the overall performance due to the linearizing effect
that local feedback around the valve has.
Goodwin,
Chapter 10
Graebe, Salgado, Prentice Hall 2000
Figure 10.7: Example of application of cascade control
r(t)
ym (t)
r(t)
ym (t)
+
C(s)
+
C1 (s)
qm (t)
+
C2 (s)
u(t)
u(t)
supply
pressure
ps (t)
a)
q(t)
Non-cascade
Valve Controller
ps (t)
b)
Cascade
Valve Controller
q(t)
Goodwin,
Chapter 10
Graebe, Salgado, Prentice Hall 2000
Figure 10.8: Cascade control structure
The generalization of this idea has the structure as
shown below:
Go2 (s)
Dg (s)
R(s)
+
C1 (s)
U1 (s)
+
C2 (s)
Go1 (s)
+
+
Inner cascade loop
Outer loop
Ga (s)
Gb (s)
Y (s)
Chapter 10
Goodwin,
Graebe, Salgado, Prentice Hall 2000
Referring to Figure 10.8 (previous slide), the main
benefits of cascade control are obtained
(i) when Ga(s) contains significant nonlinearities that limit
the loop performance;
or
(ii) when Gb(s) limits the bandwidth in a basic control
architecture.
Goodwin,
Chapter 10
Graebe, Salgado, Prentice Hall 2000
Example of Cadcade Control
Consider a plant having the same nominal model as
in the previous example on disturbance feedforward.
Assume that the measurement for the secondary loop
is the input to G02(s),
1
Go1 (s) =
;
s+1
es
Go2 (s) =
;
2s + 1
Ga (s) = 1;
es
Gb (s) = Go2 (s) =
2s + 1
We first choose the secondary controller to be a PI
controller where
C2 (s) =
8(s + 1)
s
Goodwin,
Chapter 10
Graebe, Salgado, Prentice Hall 2000
This leads to an inner loop having effective closed
loop transfer function of
8
To2 (s) =
s+8
Hence the primary (or outer loop) controller sees an
equivalent plant with transfer function
8es
Goeq (s) = 2
2s + 17s + 8
The outter controller is then designed using a Smith
Predictor (see Chapter 7).
Chapter 10
Goodwin,
Graebe, Salgado, Prentice Hall 2000
The results for the same disturbance as in the
earlier example on disturbance feedforward are
shown in the next slide. [A unit step reference is
applied at t = 1 followed by a unit step disturbance
at t = 5].
Goodwin,
Chapter 10
Graebe, Salgado, Prentice Hall 2000
Figure 10.9: Disturbance rejection with a cascade
control loop
1.4
Plant response
1.2
1
0.8
0.6
0.4
0.2
0
10
Time [s]
12
14
16
18
20
Chapter 10
Goodwin,
Graebe, Salgado, Prentice Hall 2000
Comparing Figure 10.9 with Figure 10.3 we see that
cascade control has achieved similar disturbance
rejection (for this example) as was achieved earlier
using disturbance feedforward.
Chapter 10
Goodwin,
Graebe, Salgado, Prentice Hall 2000
The main features of cascade control are
(i) Cascade control is a feedback strategy.
(ii) A second measurement of a process variable is required.
However, the disturbance itself does not need to be measured.
Indeed, the secondary loop can be interpreted as having an
observer to estimate the disturbance.
(iii) Measurement noise in the secondary loop must be considered
in the design, since it may limit the achievable bandwidth in
this loop.
(iv) Although cascade control (in common with feedforward)
requires inversion, it can be made less sensitive to modeling
errors by using the advantages of feedback.
