基于排列熵和变量预测模型的旋转机械故障诊断  

Rotating machinery fault diagnosis based on permutation entropy and variable predictive model

作　　者：裴红蕾[1] PEI Honglei(Wuxi Institute of Art and Technology,Yixing 214200,CHN)

机构地区：[1]无锡工艺职业技术学院,江苏宜兴214200

出　　处：《制造技术与机床》2018年第8期92-96,共5页Manufacturing Technology & Machine Tool

基　　金：国家自然科学基金资助项目(51105323)

摘　　要：为了提高旋转机械故障诊断的性能,提出了基于排列熵和变量预测模型的故障诊断方法。使用固有时间尺度分解法分解初始故障信号,得到多个固有旋转分量;选择含有故障信息的分量并计算其排列熵,作为故障诊断的特征值;使用训练样本得到的特征值训练不同故障状态下每个特征值的预测模型,将此变量预测模型用于诊断故障信号的故障类型;经实验验证,与BP神经网络、支持向量机等方法相比,所提出的故障诊断方法不仅诊断准确率高,而且耗时少,说明该方法有效性。To improve property of rotating machinery fault diagnosis,fault diagnosis method based on permutation entropy and variable predictive model is proposed. Intrinsic time-scale decomposition is used to decompose initial fault signal,and some proper rotation component is got. The component containing fault information is chosen,and calculates its permutation entropy as eigenvalue. These eigenvalues are used to train variable prediction model of every eigenvalue in every state,and the prediction model is used to diagnose fault type. By trial,compared with BP neutral network and support vector machine,diagnosis precision rate is higher and time host is shorter by the method designed by this paper,which can verify validity of the fault diagnosis method.

关 键 词：旋转机械 故障诊断 排列熵 变量预测模型 固有时间尺度分解 

分 类 号：TH165.3[机械工程—机械制造及自动化] TH17