Bearing fault diagnosis method based on ICEEMDAN and Fast spectral kurtosis improved by wavelet packet

A bearing feature extraction method based on ICEEMDAN and wavelet packet improved fast spectral kurtosis method is proposed to address the problem of pattern aliasing in bearing fault feature extraction and the shortcomings of the wavelet packet method in frequency band division. Firstly, the original vibration signal is processed using ICEEMDAN and decomposed into multiple intrinsic mode functions (IMFs); Select out IMF components that meet the criteria of variance contribution rate and correlation coefficient, perform signal reconstruction, perform improved wavelet packet decomposition on the reconstructed signal, and calculate the improved fast spectral kurtosis based on the kurtosis criterion to obtain the improved fast spectral kurtosis graph. Furthermore, the envelope spectrum is obtained to determine the fault feature information of rolling bearings. The feasibility of this method was verified through simulation signals and experimental analysis of rolling bearings.