基于PCA-LMD的滚动轴承信号混合特征选取与故障诊断  

Hybrid Feature Selection and Fault Diagnosis of Rolling Bearing Signals Based on Principal ComponentAnalysis Denoising and Local Mean Decomposition(PCA-LMD)

作　　者：朱天煦 臧朝平[1] ZHU Tian-xu;ZANG Chao-ping(College of Energy and Power Engineering,Nanjing University of Aeronautics&Astronautics,Nanjing 210016,China)

机构地区：[1]南京航空航天大学能源与动力学院,南京210016

出　　处：《航空发动机》2020年第5期14-21,共8页Aeroengine

基　　金：国家自然科学基金委员会与中国工程物理研究院联合基金(NO.U1730129)资助。

摘　　要：为了对非平稳、低信噪比的轴承振动信号进行分析,提出1种基于PCA-LMD的滚动轴承振动信号混合特征选取及智能故障诊断方法。基于Hankel矩阵对实测轴承振动信号进行主成分分析(PCA)降噪处理。对降噪后的非平稳信号进行局部均值分解(LMD),得到一系列具有瞬时物理意义的乘积函数(PF)。通过特征分析和对比,选取前5阶PF分量的能量比特征、样本熵、均方根及波形指标作为信号混合特征向量。将特征向量输入到支持向量机(SVM)分类器进行训练与测试,从而实现故障诊断。结果表明:通过对包含不同故障程度的滚动体、内圈、外圈故障的轴承实测数据进行分析,故障诊断正确率达到98%,验证了本方法的有效性,对航空发动机轴承的故障诊断具有借鉴和指导作用。In order to analyze the bearing vibration signals with nonstationarity and low signal-to-noise ratio,a hybrid feature selection and intelligent fault diagnosis method of rolling bearing vibration signal were proposed based on PCA-LMD.Based on the Hankel matrix,the Principal Component Analysis(PCA)and denoising of the measured bearing vibration signal were carried out.A series of Product Functions(PF)with instantaneous physical meaning were obtained by Local Mean Decomposition(LMD)of nonstationary signal after denoising.According to the analysis and comparison of features,the energy ratio feature,sample entropy,root mean square and waveform index of the first 5 PF components were selected as the hybrid eigenvectors of the signal.The eigenvectors were input into Support Vector Machine(SVM)classifier for training and testing to realize fault diagnosis.The results show that the correct rate of fault diagnosis is 98%by analyzing the measured data of rolling body,inner ring and outer ring which contain different fault degree.The validity of this method is verified,and it can be used for reference and guidance for the fault diagnosis of aeroengine bearings.

关 键 词：混合特征选取 主成分分析 降噪 局部均值分解 支持向量机 航空发动机 

分 类 号：V214.33[航空宇航科学与技术—航空宇航推进理论与工程]