基于多源信息融合的往复式压缩机故障诊断方法  

Reciprocating Compressor Fault Diagnosis Technology Based on Multi-source Information Fusion

作　　者：张明[1] 江志农[1] 

机构地区：[1]北京化工大学诊断与自愈研究中心,北京100029

出　　处：《机械工程学报》2017年第23期46-52,共7页Journal of Mechanical Engineering

基　　金：国家重点基础研究发展计划(973计划;2012CB026000);国家自然科学重点基金(51135001)资助项目

摘　　要：往复式压缩机结构复杂,振动激励源多,故障关联性较强,需要依靠多种类型的传感器所采集的信息来对往复式压缩机故障进行诊断。在融合往复式压缩机多种类型传感器采集的特征信息基础上,提出一种基于多源信息融合的往复式压缩机故障诊断方法,构建信息融合诊断框架。利用往复式压缩机多种类型传感器所采集的数据信息构建特征证据体,使用径向基神经网络对每个证据体进行初步诊断,根据加权证据融合理论融合各个证据体初步诊断结果,得到最终诊断结果。使用提出的方法对往复式压缩机3种工况的试验数据进行融合诊断,诊断结果表明:使用加权证据融合理论融合多源传感器信息的诊断结果可信度高,不确定性小,能够准确对往复式压缩机故障状态进行诊断识别。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.

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

分 类 号：TP306[自动化与计算机技术—计算机系统结构]