Bearing fault feature extraction based on optimized EMD by adaptive resonance

Abstract

According to the problem of how to select the sensitive intrinsic mode functions (IMFs), which can reflect the fault characteristics effectively, a novel sensitive IMF selected method based on adaptive resonance (AR) is proposed. Firstly, a series of IMFs are obtained by EMD, and the power spectrums are obtained for the original vibration signal and each IMF component. Then the resonant frequencies are identified, and the resonant frequencies are taken as the center frequencies of the band-pass filters. Secondly, the band-pass filters are designed to obtain the filtering frequency bands of original vibration signal and each IMF. And the subordinate relationship between filtering band of each IMF with the filtering band of the original vibration signal are compared to select the sensitive IMF. Finally, Perform envelope demodulation analysis to the IMF, and the characteristic information are provided for the subsequent fault diagnosis. The method is applied to the simulated signal and the actual rolling bearing ball fault signal, and they all verify the effectiveness of the method, and it is superior to that select sensitive IMF based on the kurtosis maximum principle.