基于EMD信息熵和支持向量机的往复压缩机轴承故障诊断分析

The fault diagnosis of reciprocating compressor bearing based on EMD information entropy and support vector machine

为了延长往复压缩机轴承的使用寿命,以实现降低压缩机故障频率的目的,采用经验模态分解(EMD)方法对压缩机轴承信号进行分解,得出若干本征模函数(IMF)分量,并利用支持向量机(SVM)分类器对其进行识别,以提高压缩机故障诊断效率。结果表明:经EMD分解得到的信号带宽是从低频到高频不断进行变化的,可以更好地突出信号局部特征;利用参数寻优函数SVMcgForRegress.m进行参数优化,最终确定最佳组合为c=0.2和g=2.624 9;SVM网络可对大头和小头轴瓦间隙大等故障类型进行识别,准确率高达96%,由此证明了SVM分类器的可靠性,同时也说明其具备良好的识别能力。

In order to extend the service life of reciprocating compressor bearings and implement the purpose of reducing the compressor failure frequency,the empirical mode decomposition（EMD） method was used to decompose the compressor bearing signals to obtain a number of IMF components,and support vector machine（SVM） classifierwas used to identify them to improve the efficiency of the compressor fault diagnosis.The results showed that the signal bandwidth obtained from the EMD decomposition is constantly changing from low frequency to high frequency,which could better highlight the local characteristics of the signal.Parameter optimization was performed by using the parameter optimization function SVMcg ForRegress.m,and finally the optimal combination was determined as c = 0.2 and g = 2.624 9; SVM network could identify the types of fault such as the gap between head and small head bearing,with the accuracy of 96%,which proved the reliability and good recognition ability of SVM classifier.