A new scheme for fuzzy rule-based system identification and itsapplication to self-tuning fuzzy controllers

摘要

There are many important issues that need to be resolved fornidentification of a fuzzy rule-based system using clustering. We addressnthree such important issues: 1) deciding on the proper domain(s) ofnclustering; 2) deciding on the number of rules; and 3) getting anninitial estimate of parameters of the fuzzy systems. We justify that onenshould start with separate clustering of X (input) and Y (output). Wenpropose a scheme to establish correspondence between the clustersnobtained in X and Y. The correspondence dictates whether furthernsplitting/merging of clusters is needed or not. If X and Y do notnexhibit strong cluster substructures, then again clustering of X* (inputndata augmented by the output data) exploiting the results of separatenclustering of X and Y, and of the correspondence scheme is recommended.nWe justify that usual cluster validity indices are not suitable fornfinding the number of rules, and the proposed scheme does not use anyncluster validity index. Three methods are suggested to get the initialnestimate of membership functions (MFs). The proposed scheme is used tonidentify the rule base needed to realize a self-tuning fuzzy PI-typencontroller and its performance is found to be quite satisfactory