An optimal frequency band selection approach via periodic shock indicator for bearing fault feature extraction

Abstract There are two challenges in the bearing fault feature extraction approach based on optimal frequency band (OFB) selection. The first is to design a high-precision signal decomposition algorithm to ensure that no fault information is lost; The second is to construct an indicator that is sensitive to the statistical characteristics of fault information. This paper proposed a novel OFB selection approach to solve the above challenges. Firstly, a parameter selection strategy is introduced for fixed bandwidth overlap-and-slip filter banks (FBOSFB). Adopting this strategy, the FBOSFB can decompose the signal into several narrowband signals while ensuring that no fault information is lost. Secondly, a novel indicator, namely periodic shock indicator (PSI), is constructed. The PSI has the capability to simultaneously assess both the periodic and shock characteristics of the bearing fault feature signal. Thirdly, the narrowband signals obtained from the first step are used as input to the OFB search algorithm, and then the OFB is determined by maximizing the PSI. Finally, three real bearing fault data are employed to verify the fault feature extraction performance of the proposed approach, the results show that the proposed approach can effectively extract the bearing fault feature.