基于LMD和样本熵的齿轮故障特征提取方法研究  

Research on Gear Fault Feature Extraction Based on LMD and Sample Entropy

作　　者：王洪明[1] 郝旺身[1] 韩捷[1] 董辛旻[1] 郝伟[1] 欧阳贺龙 

机构地区：[1]郑州大学机械工程学院,河南郑州450001

出　　处：《郑州大学学报（工学版）》2015年第3期44-48,共5页Journal of Zhengzhou University（Engineering Science）

基　　金：河南省教育厅科学技术研究重点项目(13A460673)

摘　　要：针对齿轮故障信号的非线性、非平稳特征,采用局部均值分解(LMD)结合样本熵的方法提取故障特征.采用滑动平均法构造均值函数与包络函数,将原始信号分解得到一系列的PF分量,通过剔出无意义的PF分量,筛选出反映真实状态信息的分量,然后计算筛选出的PF分量的样本熵.不同故障信号的PF分量的样本熵的大小不一,规律可寻,据此可以将样本熵的值作为元素构造故障特征向量.通过实验模拟齿轮正常、齿根裂纹、断齿和缺齿这4种状态,比较LMD-近似熵与LMD-样本熵的分类效果,实验模拟表明:LMD-样本熵比LMD-近似熵有更好的区分效果.For the non-linear and the non-stationary characteristics of gear faults signal, this study adopts the local mean decomposition （LMD） combined with the sample entropy method to extract fault features. With the moving average method to construct the mean function and the envelope function, the original signal is decomposed into a series of components PF. Then by eliminating the meaningless components so that the components including real status information could be selected to calculate sample entropy. The sample entropy changed regularly with different fault signals＇ PF, and accordingly the sample entropy could be used as elements of fault feature vector. Through experiments simulated under gear normal, tooth root cracked, tooth broken and missing teeth conditions, then compared the classification results of LMD-approximate entropy with LMD-sampie entropy, and eventually it is proved that the LMD-sample entropy is better than the LMD-approximate entropy in distinguishing these four typical conditions.

关 键 词：非线性 LMD 样本熵 故障特征 齿轮 

分 类 号：TH133[机械工程—机械制造及自动化]