A Two-Stage Rolling Bearing Weak Fault Feature Extraction Method Combining Adaptive Morphological Filter with Frequency Band Selection Strategy

Abstract

To extract the weak fault features hidden in strong background interference in the event of the early failure of rolling bearings, a two-stage based method is proposed. The broadband noise elimination ability of an adaptive morphological filter (AMF) and the superior capability of a frequency band selection (FBS) strategy for fault transient location identification are comprehensively utilized by the proposed method. Firstly, the AMF with a simple theory and high calculation efficiency is used as a preprocessing program to enhance the fault transient features. Then, the proposed FBS strategy based on the sparsity index (SI) is utilized to further handle the filtered signal processed by the AMF. Finally, the constructed optimum bandpass filter based on the analysis result of the FBS is used to further filter the handled signal processed by AMF and envelope spectral analysis is applied on the last filtered signal to realize the ideal fault feature extraction effect. Compared with the other traditional FBS methods based on kurtosis or the other index, the proposed FBS strategy based on SI has strong robustness to noise. One experimental signal and one engineering vibration signal are used, respectively, to verify the feasibility of the proposed method.