局部鲁棒主成分分析及其在故障诊断中的应用  

Local Robust Principal Component Analysis and Its Application in Fault Diagnosis

作　　者：苏立鹏 金樟民 尤戈 易灿灿[2] SU Li-peng;JIN Zhang-min;YOU Ge;YI Can-can(Wenzhou Special Equipment Inspection and Research Institute,Zhejiang Wenzhou 325000,China;Wuhan University of Science and Technology,Hubei Wuhan 430081,China)

机构地区：[1]温州市特种设备检测研究院,浙江温州325000 [2]武汉科技大学,湖北武汉430081

出　　处：《机械设计与制造》2021年第8期246-249,255,共5页Machinery Design & Manufacture

基　　金：国家自然科学基金(51805382);浙江省级质监科研项目(20180363)。

摘　　要：机械设备故障振动信号的分析一般需要经过特征提取,然而由于背景噪声或者环境干扰的存在使得信号的信息适用性下降,从而导致特征提取存在很大的困难。一种新的局部鲁棒主成分分析的降噪方法被提出,该方法假设数据矩阵在有限个局部区域可以分解为表示信号特征信息的低秩成分和代表噪声的稀疏成分的加权和,且矩阵只需在局部区域具有低秩的属性而不必要满足全局低秩的强条件,并通过有限个局部低秩矩阵的平滑凸组合来全局逼近原始矩阵。通过仿真实验和实测的轴承外圈故障数据的分析,证明了提出的方法具有较强的降噪和特征提取效果。Analyzing the fault vibration signals of mechanical equipment usually requires feature extraction.However,the information applicability of signals may be reduced and the feature extraction may be very difficult due to the background noise or environmental interference.A new de-noise method called local robust principal component Analysis is proposed.The method assumes that the data matrix can be decomposed into the weighted sum of the low-rank components representing the signal feature information in a finite local region and the sparse components representing.It is proposed that the matrix only needs to have low-rank properties in the local region instead of satisfying the strong condition of global low-rank.And the original matrix is approximated by a smoothed convex combination of low-rank matrices.The analyzing of numerical simulation experiments and the measured bearing outer ring fault signal indicates the proposed method has strong performance of noise reduction and feature extraction.

关 键 词：机械设备故障诊断 特征提取 局部鲁棒主成分分析 平滑核函数 

分 类 号：TH16[机械工程—机械制造及自动化] U462.1[机械工程—车辆工程]