基于KL-HVD的转子振动故障诊断方法研究  

A study on the method of rotor vibration fault diagnosis based on KL-HVD

作　　者：朱霄珣[1] 周沛 苑一鸣 徐博超 韩中合[1] ZHU Xiaoxun;ZHOU Pei;YUAN Yiming;XU Bochao;HAN Zhonghe(Department of Power Engineering,North China Electric Power University,Baoding 071003,China)

机构地区：[1]华北电力大学动力工程系,河北保定071003

出　　处：《振动与冲击》2018年第16期249-255,共7页Journal of Vibration and Shock

基　　金：国家科技支撑计划(2014BAA06B00)

摘　　要：希尔伯特振动分解(Hilbert Vibration Decomposition,HVD)由于其虚假分量问题,严重制约了其在实际故障诊断中的应用。针对该问题,引入信息论中的K-L散度概念,提出了基于K-L散度的HVD虚假分量识别方法(KL-HVD)。KL-HVD将HVD分量视作概率分布各不相同的信号,并且认为真实分量与原信号的概率分布较为相近。该方法在原HVD方法基础上,计算HVD各分量与原信号的K-L散度值,对分量的虚假性进行量化。由于真假分量之间具有较大的差异性,选用高斯混合模型对这些分量进行聚类,自动区分出虚假分量并予以去除。此外,分别利用互信息及相关系数方法对虚假分量问题进行研究。并将三种方法应用于转子振动问题分析,结果显示三者中KL-HVD方法能够更有效地识别虚假分量,更清晰地提取出故障的时频特征。The false components of HVD have seriously restricted its application in practical fault diagnosis.To solve this problem,Kullback-leibler(K-L)divergence was employed,which is the concept of information theory with HVD(KL-HVD).Components of HVD were treated as signals with various probability distributions,supposing that the probability distributions of real components are similar to the original signal.The false component identification method is based on K-L divergence(KL-HVD).KL-HVD,used in K-L divergence as a distinguishing index,was proposed to solve this problem.Based on the original HVD method,KL-HVD first calculates the K-L divergence values between the HVD components and the original signal and then clustered these values according to the Gaussian mixture model.Finally,the trues and the falses could automatically be separated from each other because of their intrinsic differences,and the false components will be eliminated.The results of the rotor fault signal analysis verify that KL-HVD divergence is more suitable for identifying the HVD false components than mutual information and the correlation coefficient method,and it could extract the faults’time-frequency characteristics more clearly.

关 键 词：希尔伯特振动分解 虚假分量 K-L散度 高斯混合模型 振动故障诊断 

分 类 号：TK267[动力工程及工程热物理—动力机械及工程]