RI 9515 - Aspects of Genetic Algorithm-Designed Fuzzy Logic Controllers

U.S. Bureau of Mines (USBM) research has shown that fuzzy logic and genetic algorithms (GA's) have great potential for industrial process control. A technique has been developed by the USBM in which GA's are used to substantially reduce the time needed to design fuzzy logic controllers (FLC's).This technique shows promise as an efficient, robust approach to developing control systems. The research described in this report is twofold. First, the basic approach to developing an FLC using GA's is presented. The GA-designed FLC is developed for a specific physical system, a pH titration system. Second, empirical results are presented in which variations in the FLC implementation are compared. Specifically, the effects of altering five aspects of the pH FLC are considered: (1) membership function form, (2) number of fuzzy classes, (3) the center-of-area method, (4) implication operator, and (5) fuzzy rule form. Results indicate that (1) the technique in which GA's are used to design FLC's is effective, and (2) when this technique is used, variations in FLC implementation have little effect on FLC performance in a chosen control problem.