Documente Academic
Documente Profesional
Documente Cultură
TABLE I
FUZZY RULES FOR COMPUTATION OF 4u
B. Scaling Factors
The use of a normalized domain requires input normalization,
which maps the physical values of the process state variables
into a normalized domain. In addition, output denormalization
Fig. 4. Performance measures of step response.
maps the normalized value of the control output variable into its
physical domain. The scaling factors which describe the partic-
ular input normalization and output denormalization play a role
similar to that of the gains of a conventional controller. Hence,
they are of utmost importance with respect to controller stability
and performance. They are the source of possible instabilities,
oscillation problems and deteriorated damping effects [15].
The relationship between the scaling factors
and the input and output variables of the FLC
is , , .
Adjusting the scaling factors can alter the corresponding
regions of the fuzzy sets. For example, an error equal to 0.1
may belong to PS more than to ZE as its scaling factor is
increased. Selection of suitable values of , ,
Fig. 5. Performance of PI-like FLC with constant output scaling factor at
are made based on expert knowledge about the process to =
setpoint 1 and 5 V.
be controlled, and through trial and error. Adjustment rules
have been developed for the scaling factors by evaluating
1 V, 2 V, , 5 V, where 1 V corresponds to a belt speed of
control results (e.g., the characteristics of the step response
5.08 10 m/s (1 ft/min). Considering that the desired feedrate
and heuristics) [15], [21], [22]. The evaluation performance
of the weigh belt feeder is a constant value, controller design
measures are “overshoot”(OV), “reaching time”(RT),
for variable magnitude step references was not considered in
“amplitude”(AM), and “delay time”(L), as shown in Fig. 4.
this research. In the FLC design fixed and were chosen
The adjustment rules are good reference for manual tuning by
and their values were tuned based on the adjustment rules in
human operators. Recently, numerous papers have explored the
[15], [21], [22]. A constant output scaling factor was first used
integration of genetic algorithms or neural networks with fuzzy
for the five different setpoints, with , and
systems in so-called genetic fuzzy or neural fuzzy systems.
. Fig. 5 shows the resulting experimental results
Many publications are concerned with the design of FLCs by
at setpoints of 1 and 5 V (please refer to [1] for a detailed de-
tuning the rule bases, MFs, and scaling factors [24]–[28].
scription of the weigh belt feeder experimental system). While
It has been experimentally observed that a conventional FLC
the performance of the FLC is fine with a setpoint of 1 V, the
with constant scaling factors and a limited number of IF–THEN
FLC leads to increasingly large overshoots as the setpoint in-
rules may have limited performance for a highly nonlinear plant.
creased. Thus, the proposed FLC with a constant output scaling
(This was found to be true for the weigh belt feeder.) As a result,
factor for different setpoints had degraded performance at the
there has been significant research on tuning of FLCs where
higher setpoints due to the nonlinearity of the feeder. To remedy
either the input or output scaling factors or the definitions of the
this problem, reduced control effort is needed for the higher
MFs and sometimes the control rules are tuned to achieve the
setpoints.
desired control objectives [11], [13], [18], [22]. In the following,
Based on the aforementioned observations, we proposed
we concentrate only on the tuning of output scaling factor due
tuning the scaling factor by gain scheduling it at different
to its strong influence on the performance and stability of the
setpoints. The design algorithm uses a coefficient to adjust
system.
as follows:
C. Gain Scheduling of the Fuzzy PI-Like Controller
(1)
The weigh belt feeder is normally operated within a setpoint
range of V, and 5 V is the maximum possible value of the where stands for setpoint and is some reference value
reference command. Controllers were designed for setpoints of of . The value of is determined by ,
314 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 11, NO. 3, JUNE 2003
TABLE II
FUZZY RULES FOR COMPUTATION OF
TABLE IV
FUZZY RULES FOR COMPUTATION OF K
TABLE V
FUZZY RULES FOR COMPUTATION OF T
[14] J. Carvajal, G. Chen, and H. Ogmen, “Fuzzy PID controller: design per- [30] R. Krishnan, Electric Motor Drives. Upper Saddle River, NJ: Prentice-
formance evaluation, and stability analysis,” Inform. Sci., vol. 123, no. Hall, 2001.
3, pp. 249–270, 2000.
[15] D. Driankov, H. Hellendoorn, and M. Reinfrank, An Introduction to
Fuzzy Control. Berlin, Germany: Springer-Verlag, 1993.
[16] Z. Z. Zhao, M. Tomizuka, and S. Isaka, “Fuzzy gain scheduling of PID
controller,” IEEE Trans. Syst., Man, Cybern., vol. 23, pp. 1392–1398,
Oct. 1993. Yanan Zhao received the B.S. and M.S. degrees from
[17] S. Z. He, S. Tan, and F. L. Xu, “Fuzzy self-tuning of PID controller,” Beijing Institute of Technology, Beijing, China, and
Fuzzy Sets Syst., vol. 56, no. 1, pp. 37–46, 1993. the Ph.D degree in mechanical engineering from the
[18] R. Bandyopadhyay and D. Patranabis, “A fuzzy logic based PI con- Florida State University, Tallahassee, in 1987, 1990,
troller,” ISA Trans., vol. 37, no. 3, pp. 227–235, 1998. and 2001, respectively.
[19] J. H. Lee, “On methods for improving performance of PI-type fuzzy Her professional interests include intelligent
logic controllers,” IEEE Trans. Fuzzy Syst., vol. 1, pp. 298–301, Aug. control systems for autonomous vehicles, automated
1993. controller tuning, system identification, numerical
[20] H. X. Li, “Enhanced methods of fuzzy logic control,” in Proc. IEEE Int. optimization, and modeling, simulation, and analysis
Conf. Fuzzy Systems, Yokohama, Japan, Mar. 1995, pp. 331–336. of dynamic system.
[21] W. C. Daugherity, B. Rathakrishnan, and J. Yen, “Performance evalua-
tion of a self-tuning fuzzy controller,” in Proc. IEEE Int. Conf. Fuzzy
Systems, San Diego, CA, Mar. 1992, pp. 389–397.
[22] M. Maedo and S. Murakami, “A self-tuning fuzzy controller,” Fuzzy Sets
Syst., vol. 51, no. 1, pp. 29–40, 1992.
[23] Y. P. Singh, “A modified self-organizing controller for real-time process Emmanuel G. Collins, Jr. received the Ph.D. degree
control application,” Fuzzy Sets Syst., vol. 96, no. 2, pp. 147–159, 1998. in aeronautics and astronautics from Purdue Univer-
[24] J. R. Jang, “ANFIS: adaptive-network-based fuzzy inference system,” sity, West Lafayette, IN, in 1987.
IEEE Trans. Syst., Man, Cybern., vol. 23, pp. 665–685, June 1993. He worked for seven years in the Controls Tech-
[25] A. Nurnberger, D. Nauck, and R. Kruse, “Neuro-fuzzy control based on nology Group at Harris Corporation, Melbourne,
the NEFCON-model: recent development,” Soft Comput., vol. 2, no. 4, FL, before joining the Department of Mechanical
pp. 168–182, 1999. Engineering at the Florida A & M University,
[26] A. Arslan and M. Kaya, “Determination of fuzzy logic membership Florida State University College of Engineering,
functions using genetic algorithms,” Fuzzy Sets Syst., vol. 118, no. 2, Tallahassee, where he currently serves as Professor.
pp. 297–306, 2001. His current research interests include intelligent
[27] K. Warwick and Y. H. Kang, “Self-tuning proportional, integral and control systems for autonomous vehicles, robust
derivative controller based on genetic algorithm least squares,” J. Syst. fault detection and isolation, control in manufacturing, automated controller
Control Eng., vol. 212, no. 16, pp. 437–448, 1998. tuning, automated weight selection in modern control, and fluidic thrust vector
[28] A. Homafier, M. Bikdash, and V. Gopalan, “Design using genetic algo- control. He is the author of over 180 technical publications in control and
rithms of hierarchical hybrid fuzzy-PID controllers of two-link robotic robotics.
arms,” J. Robot. Syst., vol. 14, no. 6, pp. 449–463, 1997. Dr. Collins served as an Associate Editor of the IEEE TRANSACTIONS ON
[29] K. Astrom and T. Hagglund, PID Controller: Theory, Design and CONTROL SYSTEM TECHNOLOGY from 1992 to 2001. In 1991, he received the
Tuning, 2nd ed. Research Triangle Park, NC: Instrum. Soc. Amer., Honorary Superior Accomplishment Award from the NASA Langley Research
1995. Center for “contributions in demonstrating active control of flexible spacecraft.”