Extraction of compound fault features of rolling bearing based on robust local mean decomposition and Wiener filter

Upon the occurrence of a fault in rolling bearing, vibration signals often contain a great many of noises making it difficult to extract fault features. Though Wiener filter is featured by excellent denoising effect, it is often hard to precisely determine its input signal. To solve this problem, the paper has proposed a fault feature extraction method of rolling bearing integrating robust local mean decomposition (RLMD) and Wiener filter. Firstly, original vibration acceleration signals are decomposed based on RLMD to obtain optimal fitting product function (PF); secondly, the PF component of strongest relevance to original signals is chosen according to correlation coefficient to determine input signal for Wiener filter; finally, spectral analysis is given to the self-correlation function of filtered signals, and fault feature frequency of bearing is extracted and compound faults are judged. The results of comparative analysis with other classical methods indicates that the combined method of RLMD and Wiener filter can effectively restrain the noise and accurately identify a compound fault of bearing, which further confirms the efficacy of the proposed method.