基于LMD-MFE和DHMM的滚动轴承故障诊断算法

Fault Diagnosis Algorithm of Rolling Bearings Based on LMD-MFE and DHMM

针对滚动轴承在恶劣环境影响下,其特征信息难以被有效提取出来的问题,提出一种基于局部均值分解(Local Mean Decomposition,LMD)和多尺度模糊熵(Multiscale Fuzzy Entropy,MFE)的滚动轴承故障诊断算法。首先,利用LMD对轴承振动信号进行分解,得到一系列乘积函数(Product function,PF)分量,并与经验模态分解(Empirical Mode Decomposition,EMD)进行对比,分析其优越性;然后对每一个分量分别提取MFE特征,同时与多尺度样本熵(Multiscale Sample Entropy,MSE)进行对比,分析MFE的优越性;最后结合各个轴承状态的类间平均距离对多个尺度因子下的熵值进行优选,筛选出可分性良好的敏感特征集,并输入到离散隐马尔科夫模型(Discrete Hidden Markov Models,DHMM)模式分类器中对轴承故障类型进行诊断识别。实验结果表明,所提出的基于LMD和MFE的轴承故障诊断算法能较好识别出多种轴承故障类型。

In poor environment,the rolling bearing characteristic information can hardly be extracted effectively.In this paper,a rolling bearing fault diagnosis method based on local mean decomposition(LMD)and multi-scale fuzzy entropy(MFE)is proposed.Firstly,LMD is used to decompose the vibration signals into a series of product function(PF)components and the result is compared with that of the empirical mode decomposition(EMD).The superiority of this method is analyzed.Then,the MFE of each component is extracted and compared with the multi-scale sample entropy(MSE).And the superiority of the MFE method is analyzed.Finally,the best sensitive feature sets are extracted based on the maximum value of average distances between different states.The best sensitive feature sets are input into discrete hidden Markov models(DHMM)and the fault identification of the bearing condition is completed.Experimental study has shown that this bearing diagnosis algorithm can effectively extract the bearing fault feature information.