基于混合域特征优选的电机轴承故障诊断  

Motor Bearing Fault Diagnosis Based Hybrid Domain Feature Optimization

作　　者：胡文浩 吴金龙 董建林 HU Wen-hao;WU Jin-long;DONG Jian-lin(School of Mechanical&Automotive Engineering,Qingdao University of Technology,Qingdao 266520,China)

机构地区：[1]青岛理工大学机械与汽车工程学院,山东青岛266520

出　　处：《机械工程与自动化》2024年第4期32-35,共4页Mechanical Engineering & Automation

摘　　要：为解决混合域特征维数高且存在冗余特征从而造成电机轴承故障诊断精度不高的问题,提出了一种基于混合域特征优选的电机轴承故障诊断方法。首先,利用自适应噪声完备集合经验模态分解处理信号,提取分解得到的前4个固有模态函数分量的熵特征和重构信号的时域、频域特征构建混合域特征集;然后,采用最大相关最小冗余和随机森林相结合的特征选择算法对提取的混合域特征集进行重要性排序得到特征子集;最后,将特征子集输入到利用灰狼算法优化的极限梯度提升算法进行故障诊断,并得出最优特征子集。实验结果表明:相较于单一域和混合域特征诊断方法,基于混合域特征优选的电机轴承故障诊断方法输入的特征数量更少且故障诊断准确率更高。In order to solve the problem of high dimensionality of hybrid domain features and the existence of redundant features,which results in low accuracy of motor bearing fault diagnosis,a motor bearing fault diagnosis method based on hybrid domain feature preference is proposed.Firstly,the signal is processed by using the CEEMDAN(complete ensemble empirical model decomposition adaptive noise)to extract the entropy features of the first four IMF(intrinsic modal function)components obtained from the decomposition and the time and frequency domain features of the reconstructed signal to construct the hybrid domain feature set;then,the feature selection algorithm(mRMR-RF)combining mRMR(maximum correlation and minimum redundancy)and RF(random forest)is used to rank the importance of the extracted hybrid domain feature set to obtain the feature subset.Finally,the feature subset is input into the XGBoost(extreme gradient boosting)algorithm optimized using the GWO(grey wolf algorithm)for fault diagnosis,and the optimal feature subset is derived.The experimental results show that this method has fewer input features and higher fault diagnosis accuracy than single and mixed domain feature diagnosis methods.

关 键 词：电机轴承 故障诊断 自适应噪声完备集合经验模态分解 特征选择 极限梯度提升 

分 类 号：TH133.3[机械工程—机械制造及自动化]