A New Fuzzy-neural Networks and Its Application in Fault Diagnosis

This paper presents a new approach to fault disgnosis based on modular fuzzy neural network for nonlinear systems.Firstly,the measuring space has been divided into several subspacos by using fuzzy e-means clustering.Secondly,ac- cording to the fuzzy rule's requirement,the subspaces have been fitted by local BP network respectively,Lastly,the nonlinear system's characteristies between fault outputs and measuring inputs in different subspaces have been obtained by processing off- line learning.Testing shows the network has generalization-performance and can improve the robustness and validity in nonlinear system's diagnosis.