关键词: 小波包, 模糊神经网络, 故障诊断, 数据融合, D-S证据理论

Abstract: According to the characteristics of fault diagnosis for centrifugal pump, wavelet package decomposition and reconstruction technique is used to extracting frequency band energy feature, a fusion diagnosis method is presented by using fuzzy neural network and D-S evidence theory for centrifugal pump. Firstly, according to the method of wavelet package, vibration signals of displacement and acceleration of centrifugal pump were disposed and transformed into feature vector value; then, two subFNN structures were established, and their training samples were obtained. After two subFNN were trained successfully, the intermediate diagnosis results were obtained through two subFNN. Finally, the FNN diagnosis results were used as the basic probability distribution value to each fault mode, and the D-S evidence theory was applied, and the final fusion diagnosis results were obtained. The experimental result was to verify the method presented in this paper.

Key words: wavelet packet, fuzzy neural network, fault diagnosis, data fusion, D-S evidence theory