Research on Recognition Method of Circuit Fault Diagnosis Model Based on Neural Network Fuzzy Clustering

Abstract

A model identification method is proposed, which includes both structural identification and parameter identification. In the structure identification, an unsupervised clustering algorithm is adopted, which can automatically divide the input and output space from the known input and output data without any prior assumptions about the data structure, and determine the number of fuzzy rules and the initial parameters of the premise and conclusion part of each rule. Therefore, the purpose of structure identification is to generate an initialization fuzzy model which can describe the given input and output data structure. In parameter identification, a four-layer fuzzy neural network is constructed to match the fuzzy reasoning mechanism of the fuzzy model.