关键词: 多源信息融合, 故障诊断, 加权证据理论, 径向基神经网络, 往复式压缩机

Abstract: Due to the complex structure, various vibration excitation sources and closely fault correlation, different kinds of sensor information are needed to identify faults of reciprocating compressor. Based on fused diverse kinds of sensor acquired feature information of reciprocating compressors, a method for fault diagnosis of reciprocating compressors is proposed, and a fusion diagnosis framework is constructed. Evidence feature space is constructed by using multi-sensor information of reciprocating compressors, and initially diagnosed by using RBF neural network. According to weighted evidence theory, the final diagnosis is obtained by fusing diagnostic results of the RBF neural network. Three kinds working condition of the reciprocating compressor experimental data are diagnosed by the proposed method. Diagnosis result shows that diagnosis of multi-source information fusion has high reliability and low uncertainty. The proposed method can accurately identify the reciprocating compressor fault.

Key words: fault diagnosis, information fusion, RBF neural network, reciprocating compressors, weighted evidence theory multi-source