Professional Documents
Culture Documents
1: Introduction
Ordinary fuzzy controllers have been succesfully
applied to a variety of plants since the pioneering works
of Mamdani and colaborators [1]. In the case of
multivariable processes, the natural approach would be to
consider as rules antecedents all the controller inputs,
which grow in number as the number of desired outputs,
or reference inputs, grows. This would certainly make the
process of designing the control strategy, or rule-base, a
more complex one. On the other hand, if independent
controllers are used for each reference input, rule-base
design becomes simpler and the control strategy becomes
potentially more reliable.
In the case of robot control through a learning fuzzy
controller [2], it has been shown that the use of
independent controllers for each link can give good
results. That is, the controllers, by adjusting their set of
rules, cope very well with the coupling between variables.
In this work the approach of using separate fuzzy
controllers is employed, with the purpose of verifying
whether this strategy is adequate. The plant, described in
the next section, is a fluid mixer, which, by presenting
non-linear characteristics, provides an additional
complexity and constitutes a good test for the designed
fuzzy control system.
q - q0 =
(1)
q 0 = Cd.ab.
(2)
(3)
2gh
(4)
(5)
(6)
(7)
(8)
(9)
(10)
Eh
Q
PB
PM
Z
NM
NB
Z
M
S
Z
Z
Z
S M B VB
B VB VB VB
M B VB VB
S M B VB
Z S M B
Z Z M B
Q0
Table I: Rule-base for height control
Ec
Qr NB NM Z PM PB
Z
PB NVB NB NM NS
Z
PS
PM NB NM NS
Z
PS PM
Z NM NS
Z
PS PM PB
NM NS
Z
PS PM PB PVB
NB
Ec
Table II: Rule-base for coloration control
4: Simulated experiments
The process response will depend on the scaling
factors, which can be empirically set beforehand and then
tuned in order to improve the controller performance. For
example, in the case of the fluid mixer, the range
established for q and q0 was from 0 to 60 cm3. Since the
corresponding universes of discourse are [0,4], GQ and
GQ0 were initially set at 15. This value gave origin to
satisfactory responses and did not need to be tuned. The
maximum possible height error eh was assumed to be 7
cm, which, considering the corresponding universe as
[-2,2], gives GEh=0.3. In order to have good accuracy in
coloration, and by noting that the universe of discourse is
[-2,2] for that variable, the scaling factor GEc was
0.1
0.2
0.4
0.2
0.1
Fig. 7: Experiment 1
Fig. 8: Experiment 2
Experiment 3: Coloration and height variations are shown
in Fig. 9 for different output valve openings when steps of
6: References
Fig. 9: Experiment 3
5: Conclusions
The simulated experiments have shown that the
approach of using two independent fuzzy controllers in
the control of a MIMO system can give good results. In
the plant considered in this work, the introduction of a
new variable allowed the reasoning process to occur in a
decoupled fashion.
The main objective of reaching a specified coloration
for the liquid in the mixing tank was achieved with very
good precision - error of less than 0.001. The height
controller was capable of maintaining the liquid level in
the mixing tank near the reference value, even without the
use of the change in error as a controller input.
Acknowledgement
This work was partially supported by the Brazilian
Research Agencies CAPES and CNPq.