LTSA和KECA相结合的轴承故障诊断  

Fault Diagnosis of Rolling Bearings Based on LTSA and KECA

作　　者：高胜利 党伟明 齐咏生[1] 赵小荣 GAO Sheng-li;DANG Wei-ming;QI Yong-sheng;ZHAO Xiao-rong(Institute of Electric Power,Inner Mongolia University of Technology,Inner Mongolia Huhhot 010080,China;Inner Mongolia North Longyuan Wind Power Co.Ltd.,Inner Mongolia Huhhot 010050,China;China Mobile Group Inner Mongolia Co.,Ltd.,Inner Mongolia Huhhot 010011,China)

机构地区：[1]内蒙古工业大学电力学院,内蒙古呼和浩特010080 [2]内蒙古北方龙源风力发电有限责任公司,内蒙古呼和浩特010050 [3]中国移动通信集团内蒙古有限公司,内蒙古呼和浩特010011

出　　处：《机械设计与制造》2018年第10期27-31,共5页Machinery Design & Manufacture

基　　金：国家自然科学基金项目(61364009;21466026);内蒙古自治区自然科学基金项目(2015MS0615)

摘　　要：针对轴承的工况复杂,其振动信号呈现非线性、非平稳特性。传统算法不能充分挖掘出非线性、非平稳信号内部本质信息,提出了基于局部切空间排列算法(LTSA)与核熵成份分析(KECA)相结合的故障诊断方法。该方法首先将滚动轴承振动信号一维时间序列重构到高维相空间,并估计数据的本征维数;然后利用局部切空间排列算法对数据集进行维数约简,得到初始的低维流形结构特征向量空间的第一行特征,对其进行快速傅里叶变换(FFT),从其频谱中分别提取滚动轴承内环、外环的故障特征频率及它们分别对应的倍频和频谱能量等7个变量作为故障特征向量;最后采用KECA对滚动轴承的故障特征向量进行模式识别,KECA可实现根据熵值大小进行特征分类,具有较强的非线性处理能力,从而实现故障的识别与诊断。采用Case Western Reserve大学提供的轴承实验数据对算法进行了验证,结果表明该方法可有效提取滚动轴承的故障特征,可以对滚动轴承的故障类型精确分类,实现对滚动轴承准确的故障诊断。According to the complex condition of bearing,the vibration signal is nonlinear and non-stationary.Based on traditional algorithms can not fully exploit the nonlinear and non stationary signal intrinsic information,this paper proposes based on local tangent space alignment algorithm(LTSA)and kernel entropy component analysis(KECA)combined with the method of fault diagnosis.Frist,the rolling bearing vibration signal of one-dimensional time series reconstruction to the high dimension space and estimate the intrinsic dimensionality of data,andthen use the local tangent space alignment algorithm for data sets dimension reduction and obtain low dimensional manifold structure from the original.Then get dimension reduction after feature vector space the first line features and the Fast Fourier Transform(FFT),from the spectrum were extracted from rolling bearing inner ring and outer ring of the fault characteristic frequency and which correspond to the harmonic generation and spectrum energy the seven variables as the fault feature vector.Finally,KECA is used to identify the fault feature vector of rolling bearing.The KECA nonlinear classifier is established,and a new monitoring statistic is introduced to estimate the divergence measure statistic to realize the fault diagnosis.The KECA can be realized according to the size of the entropy for feature classification,with a strong non-linear processing ability,so as to realize the fault identification and diagnosis.The Case Western Reserve University provides bearing experimental data of the algorithm was validated results show that this method can effectively extract bearing of fault feature,can of rolling bearing fault type precise classification,to achieve accurate fault diagnosis of rolling bearing.

关 键 词：滚动轴承 轴承故障诊断 局部切空间排列算法 KECA 

分 类 号：TH16[机械工程—机械制造及自动化]