集成辛奇异值模态分解及其在滚动轴承故障诊断中的应用  

Ensemble Symplectic Singular Mode Decomposition and Its Application in Fault Diagnosis of Rolling Bearings

作　　者：颜秋艳 刘玄 YAN Qiuyan;LIU Xuan(School of Automotive Engineering,Xiangtan Institute of Technology,Xiangtan 411100,Hunan,China)

机构地区：[1]湘潭理工学院汽车工程学院,湖南湘潭411100

出　　处：《噪声与振动控制》2024年第5期160-165,217,共7页Noise and Vibration Control

基　　金：2023年湖南省教育厅科学研究资助项目(23C0968);2023年湘潭市指导性科技计划资助项目(CG-ZD-JH20231002)。

摘　　要：对时间序列的分析是机械故障诊断领域中最为重要的手段,但是拾取的信号往往包含大量干扰噪声,严重影响故障诊断的准确性。因此,提出一种集成辛奇异值模态分解(Ensemble Symplectic Singular Mode Decomposition,ESSMD)降噪方法。在ESSMD方法中,采用互信息函数法和GP算法自适应设置参数,并构造辛几何相似变换矩阵,进而获得降噪分量信号。然而,通过辛几何相似变换获得的分量可能耦合噪声信息,难以通过传统的“筛选”分量进行降噪。为了弱化分量信号中的耦合噪声,在ESSMD中进一步引入拉格朗日乘子,抑制分量中噪声对纯信号信息的干扰,获得更加纯净的纯信号矩阵。仿真和实验结果表明,ESSMD能够有效减少信号中包含的噪声。The analysis of time series is the most important means in the field of mechanical fault diagnosis,but the picked up signals often contain a large amount of interference noise,which seriously affects the accuracy of fault diagnosis.Therefore,a noise reduction method called Ensemble Symplectic Singular Mode Decomposition(ESSMD)is proposed.In this method,mutual information function method and GP algorithm are used to set parameters adaptively,and the symplectic geometric similarity transformation matrices are constructed to obtain denoised component signals.However,the compo-nents obtained by symplectic geometry similarity transformation may be coupled with noise information,which is difficult to be removed through the traditional components filtering.In order to weaken the coupling noise in the component signals,the Lagrange multipliers is further introduced in ESSMD to suppress the interference of noise in the components to pure signal information and obtain a cleaner pure signal matrix.The simulation and experimental results indicate that ESSMD can effec-tively reduce the noise included in the signal.

关 键 词：故障诊断 集成辛奇异值模态分解 拉格朗日乘子 降噪 

分 类 号：TH113[机械工程—机械设计及理论]