一种ELMD模糊熵和GK聚类的轴承故障诊断方法

摘要

A roller bearing fault diagnosis method based on ensemble local mean decomposition(ELMD)fuzzy entropy and Gustafson-Kessell(GK)clustering was proposed to deal with the non-stationary and the nonlinear vibration signal form fault roller bearings.First,ELMD method was applied to decompose the vibration signals into a finite number of product function (PF)component and a residuals. Then, through the correlation analysis of PF component and original signals, PF components that have largest correlation coefficients with the vibration signal are sifted out. The fuzzy entropies of these PF components are calculated and used as the eigenvectors. Finally, the constructed eigenvectors are put into GK clustering to recognize different fault types. The analysis results from roller bearing signals with normal, inner race fault,rolling element and outer race faults show that the diagnosis method based on ELMD fuzzy entropy and Gustafson-Kessell can accurately identify roller bearing fault.%针对滚动轴承故障振动信号的非平稳、非线性特性,采用一种基于总体局部均值分解(Ensemble Local Mean Decomposition,ELMD)模糊熵和GK(Gustafson-Kessell)聚类的滚动轴承故障诊断方法.首先通过对滚动轴承故障振动信号进行ELMD分解,得到若干的乘积函数(Product Function,PF)分量和一个残差.然后,通过PF分量和原始轴承故障信号的相关性分析,选取与原始信号相关性最大的PF分量,并求取PF分量的模糊熵值作为特征向量.最终,通过GK聚类对所得的特征向量进行识别分类.通过对滚动轴承正常状态、内圈故障、滚动体故障和外圈故障的轴承四种状态分析表明,基于ELMD模糊熵和GK聚类的方法能够准确有效的对轴承故障状态进行分类识别.