Incipient feature extraction for rolling element bearing based on particle filter preprocessing and kurtogram

Abstract

Fast kurtogram (FK) has been broadly investigated on fault diagnosis for rolling element bearing(REB) since it is put forward. But its performance is low as incipient feature extraction and fault diagnosis of REB as noise has greatly effect on the classification accuracy. How to effectively improve the signal-to-noise ratio(SNR) is important for characteristic frequency(CF) determination based on FK. In this research, the SNR can be improved by using particle filter(PF) to be satisfied with noise interference. It can be useful for the improvement of FK after de-noise. Firstly, state space function for the analyzing signal is constructed, background noise of origin signal can be extracted by using the PF and improve the SNR. Then, the optimal band is chosen by using FK after PF preprocessing. In the end, fault CF can be obtained by using spectrum analysis. Simulation and monitored vibration signals for REB are used to verify the effectiveness of this method. It can be concluded that the proposed method has better performance compared with that of FK and EMD-Kurtogram on REB fault early classification.