Tuning method of linguistic membership functions

Abstract

A learning method which can tune the linguistic membership functions is presented. To accomplish such a method, the fuzzy inference mechanism must be expressed by the membership functions without losing a physical sense. The fuzzy subsets are then described by only a few membership functions. When both the control input and the linguistic membership function are tuned in the fuzzy controller, the fuzzy rules after the training can be expressed by the linguistic membership functions maintaining their physical meaning.<>