Fault Diagnosis System for Rotary Machine Based on Fuzzy Neural Networks

This paper is concerned with the application of fuzzy neural networks to fault diagnosis systems for rotary machines. In practical fault diagnosis, it is very difficult to improve the recognition rate of pattern recognition, especially when the sample data are similar. To solve these difficulties, a fault diagnosis system using fuzzy neural networks is proposed in this research. A fault diagnosis system with fuzzy neural networks is based on a series of standard fault pattern pairings between fault symptoms and fault. Fuzzy neural networks are trained to memorize these standard pattern pairs. Unlike other neural networks, fuzzy neural networks adopt bi-directional association. They make use of information from both the fault symptoms and the fault patterns, which can improve recognition rate greatly. When an unknown sample becomes the input for a trained fault diagnosis system, the fault diagnosis system can make fault diagnosis by bi-directional association of fuzzy neural networks. Through experiments with a rotor testing table and applications in monitoring and fault diagnosis of water pump sets of oil plant, it is verified that fuzzy neural networks have a well distinguished ability and are effective to perform fault diagnosis of rotary machines.