ABSTRACT
The method and strategy of combining traditional neural networks with fuzzy analysis technique, which discussed deeply in this paper, are valuable for application of solving the problems of diagnostic reasoning for a variety of mechanical faults. The fuzzy fault diagnosis put into practice by fuzzy processing of input-output of the conventional structure of neural networks. The network weights are trained and optimized by using genetic algorithm. The results of experiment performed on gear couples of NF125 type motor engine demonstrate that the hybrid system of fuzzy neural networks has a good ability of recognizing the states of samples and validity of diagnostic reasoning for border fuzzy situation of fault feature data.
