Sequential self-reorganization method of symptom parameters and identification method of membership function for fuzzy diagnosis

When building up a fuzzy diagnosis system, symptom parameters (SPs) must be extracted and the membership functions between the symptom parameters and failure categories must be defined for fuzzy inference. Currently, however, there is no acceptable method for extracting the optimum SP by which the failure types can be sensitively distinguished. In order to overcome this difficulty and ensure highly accurate failure diagnosis, in this paper, a new method called “sequential self-reorganization of symptom parameters” is proposed by using genetic algorithms (GA). Also the identification method of membership functions of symptom parameters is discussed by using the possibility theory. The efficiency of these methods is verified by applying them to a ball bearing diagnosis system. The new methods proposed here can also be applied to other pattern recognition problems