Adaptive multiple periodic impulse response mode extraction method and its application in rolling element bearing compound fault diagnosis

The coupling of multiple impulse response modes is a fundamental characteristic of vibration signals in rolling element bearing compound faults. The adaptive extraction and separation of periodic impulse response features have been recognized as a significant challenge in the compound fault diagnosis of rolling element bearings. To address this issue, an adaptive multiple periodic impulse response mode extraction method is proposed. The purpose of this method is to adaptively extract the periodic impulse response consistent with the fault characteristic frequency through the convolution of a periodic weight vector with the original signal. Periodicity-constrained kurtosis is used as the optimization objective to obtain the period of the weight vector; thus, the periodicity and impulsiveness of the fault feature are simultaneously considered. It is proved theoretically that the periodic impulse response mode extraction process is essentially complete fault harmonic extraction within the full frequency band. Adaptive mode selection is realized based on cross-correlation to extract multiple impulse response modes. The effectiveness of the proposed method is verified through theoretical analysis, as well as simulated and experimental signal analysis, and exhibits obvious superiority in bearing fault feature extraction.