基于改进奇异值分解的滚动轴承微弱故障特征提取方法  

Feature Extraction of Weak Fault for Rolling Bearing Based on Improved Singular Value Decomposition

作　　者：崔玲丽[1] 刘银行 王鑫[1] CUI Lingli;LIU Yinhang;WANG Xin(Department of Materials&Manufacturing,Beijing University of Technology,Beijing 100124)

机构地区：[1]北京工业大学材料与制造学部,北京100124

出　　处：《机械工程学报》2022年第17期156-169,共14页Journal of Mechanical Engineering

基　　金：国家自然科学基金资助项目(52075008)。

摘　　要：针对强背景噪声及谐波干扰的滚动轴承早期微弱故障特征提取,提出一种改进奇异值分解(Improved singular value decomposition,ISVD)的故障诊断新方法。首先,针对正弦信号、复合正弦信号和周期性冲击信号各自特征,根据奇异值子对(Singular value pairs,SVP)的形成原理,分别提出改进的Hankel矩阵嵌入维数优化选取原则,明确了该参数的量化范围,进而确定奇异值分解(Singular value decomposition,SVD)的最佳嵌入维数。该算法可自适应匹配SVD的Hankel矩阵最佳嵌入维数,进而获得形成SVP分布的信号分解策略。随后,结合谐波干扰的能量及SVP分布,实现对包含轴承微弱故障成分的子信号进行定位。最后,采用反对角线平均法重构目标子信号,对其进行包络谱分析获得诊断结果。仿真的滚动轴承故障信号和多组试验信号分析验证了所提方法的可行性和有效性。A novel fault diagnosis method based on improved singular value decomposition(ISVD)is proposed to extract the early weak fault feature of rolling element bearings submerged in strong background noise and harmonic interference.Firstly,according to the characteristics of sinusoidal signal,composite sinusoidal signal,periodic impact signal and the formation principle of singular value pairs(SVP),the optimization selection principles of improved embedding dimension of Hankel matrix are proposed respectively,and the quantization range of this parameter is defined.Then the optimal embedding dimension of singular value decomposition(SVD)is determined.The method can adaptively match the optimal embedding dimension of Hankel matrix of SVD.Then the signal decomposition strategy of SVP distribution is obtained.Secondly,combining with the energy of harmonic interference and SVP distribution,the sub-signals contained the weak fault information are located.Finally,the fault sub-signals are reconstructed by the inverse diagonal average method,and the diagnosis results are obtained by the envelope spectrum analysis.The feasibility and effectiveness of the proposed method are verified through the analysis results of simulated rolling bearing fault and multiple experiment signals.

关 键 词：奇异值分解 包络分析 特征提取 滚动轴承 

分 类 号：TH17[机械工程—机械制造及自动化]