齿轮局部故障的滑窗相关和重叠群稀疏诊断方法  

Sliding window correlation and overlapping group sparsity diagnosis method for gear localized fault

作　　者：吴芳坦 林慧斌[1] 何国林[1] WU Fang-tan;LIN Hui-bin;HE Guo-lin(School of Mechanical and Automotive Engineering,South China University of Technology,Guangzhou 510640,China)

机构地区：[1]华南理工大学机械与汽车工程学院,广东广州510640

出　　处：《振动工程学报》2020年第5期1084-1093,共10页Journal of Vibration Engineering

基　　金：国家自然科学基金资助项目(51875207,51875206,51705156)。

摘　　要：齿轮发生局部故障时会产生周期性冲击激励,如何在早期故障阶段或强背景噪声下提取齿轮局部故障分量一直是故障诊断的难点。针对此问题,提出一种基于滑窗相关和重叠群稀疏的齿轮局部故障特征提取方法。该方法先利用移不变K-SVD算法学习到的故障冲击模式与原始信号进行滑窗相关,对隐藏在噪声中的冲击分量进行特征增强,再利用冲击分量所具有的群稀疏特性,通过重叠群稀疏算法直接从相关信号中提取包含故障周期特征的群稀疏成分,并进一步重构出冲击信号。利用所提方法,齿轮局部故障仿真和实验信号中的故障特征和冲击分量均被很好地提取出来。此外,通过与谱峭度以及其他同类方法的对比进一步说明了所提方法的优越性,通过在正常齿轮信号上的分析也验证该方法不会产生误诊。Localized damage appears in gear may induce a series of periodic impacts,and it is a key problem to extract the localized fault component from gear vibration signal in the early failure stage or under strong background noise.In this paper,a feature extraction method based on sliding window correlation and Overlapping Group Sparsity(OGS)is proposed for gear localized fault diagnosis.The fault impact pattern learnt by shift-invariant K-SVD algorithm is used for sliding window correlation with the original signal,so that the impact feature that buried in heavy noise is enhanced in correlation signal.Using the group sparse property of periodic impacts,the fault periodic feature is extracted directly by the OGS algorithm from the correlation signal,and the impact fault signal is further reconstructed.By the proposed method,the fault feature and impact component are well extracted from the gear localized damage simulation and gearbox experimental signals.In addition,the superiority of the proposed method is further illustrated by comparison with spectral kurtosis(SK)and other similar methods.The analyzing result of the normal gear signal shows that the proposed method does not cause misdiagnosis.

关 键 词：故障诊断 齿轮 移不变K-SVD 重叠群稀疏 滑窗相关 

分 类 号：TH165.3[机械工程—机械制造及自动化] TH132.41