LMD与小波阈值降噪结合的轴承故障识别

Fault Recognition of Bearing Based on LMD and Wavelet Threshold De-noising

直接小波阈值降噪会使有用信号部分失真,为更好地对轴承故障进行特征提取,提出将局部均值分解(LMD)与小波阈值降噪结合进行降噪处理,其方法是:对含噪音的信号先进行LMD分解,并仅对分离出的高频信号分量采用改进阈值函数的小波降噪,保留残余信号的完整数据,然后重构信号。通过MATLAB仿真和轴承故障特征提取实验表明,与其它几种信号降噪方法相比,基于LMD方法并改进阈值函数的小波降噪方法,能提高信噪比,能更好的对信号进行特征提取。

The traditional wavelet threshold de-noising directly effects on the whole signal,which often causes the loss of the useful signal features. In order to effectively carry out the feature extraction for the faulted bearing,the combination of local mean decomposition( LMD) and wavelet threshold de-noising is proposed:Firstly,LMD acts on the signal mixed with noise; Wavelet de-noising based on the improved threshold function is used only for the high frequency PF component,and the complete signal data of the lowfrequency component is preserved; And then all the components are reconstructed to get the signal after de-noising.The MATLAB simulation and experiment results of the bearing fault feature extraction showthat compared with other methods of de-noising,the proposed method effectively improves signal-to-noise ratio( SNR),which is suitable for the extraction of signal features.