基于局部均值分解与拉普拉斯特征映射的滚动轴承故障诊断方法  

Fault Diagnosis Method of Bearings Based on LMD and LE

作　　者：徐倩倩[1] 刘凯[1] 侯和平[1] 徐卓飞[1] Xu Qianqian Liu Kai Hou Heping Xu Zhuofei(Xi＇an University of Technology, Xi＇an, 710048)

机构地区：[1]西安理工大学,西安710048

出　　处：《中国机械工程》2016年第22期3075-3081,共7页China Mechanical Engineering

基　　金：国家自然科学基金资助项目(51275406);国家青年科学基金资助项目(51305340)

摘　　要：针对滚动轴承非平稳振动信号的特征提取及维数优化问题,提出了融合局部均值分解与拉普拉斯特征映射的轴承故障诊断方法。首先,通过局部均值分解对非平稳振动信号进行平稳化分解,提取乘积函数分量、瞬时频率及瞬时幅值的高维信号特征集;然后,将高维特征集作为拉普拉斯特征映射算法的学习对象,提取轴承高维故障特征集的内在流形分布,以获得敏感、稳定的轴承振动特征参数,实现基于非平稳振动信号分析的滚动轴承故障特征提取;最后,结合支持向量分类模型量化LMD-LE方法的特征提取效果,实现不同状况下的轴承故障分类。轴承故障样本分类识别平均正确率达到91.17%,表明LMD-LE方法有效实现了高维局部均值分解特征集合的降噪,所提取的特征矩阵对轴承故障特征描述准确。A new diagnosis method for feature extraction of non-stationary vibration signals and fault classification of rolling bearings was proposed based on LMD and LE.Firstly,the non-stationary vibration signals of rolling bearings were decomposed into several product functions with LMD.Then,dimensional fault feature sets were established by the time-frequency domain features of product function,instantaneous frequency and amplitude.Secondly,LE was introduced to extract the sensitive and stable characteristic parameters to describe the running states of rolling bearings effectively and accurately.Finally,support vector machine classification model was built to realize the classification of fault bearings.For test samples classification,the average prediction accuracy is as 91.17%.It means that the fusion method of the LMD and LE is suitable and feasible for the bearing fault feature extraction.

关 键 词：非平稳信号 局部均值分解 拉普拉斯特征映射 故障诊断 

分 类 号：TH17[机械工程—机械制造及自动化]