Optimizing Fuzzy Membership Functions Using Particle Swarm Algorithm

The choice and shape of membership functions are known to affect the performance of fuzzy systems; despite their importance however, MFs are generally defined subjectively based on engineering judgment, designer experience or chosen for computational convenience, which does not necessarily give optimal performance when used in modeling or control. In this paper we present a method for optimizing membership functions of a fuzzy system using particle swarm optimization (PSO) algorithm. The method determines the optimal shapes and span of membership function based on a modeling performance measure. To demonstrate its effectiveness, the proposed method was used to optimize the triangular membership functions of the fuzzy model of a nonlinear system; results show that the optimized MFs provided better performance than a fuzzy model for the same system when the MFs were heuristically defined.