The feature extraction of rolling bearing fault based on wavelet packet - empirical mode decomposition and kurtosis rule

Abstract

The feature extraction method of rolling bearing fault was presented based on the combination of wavelet packet-EMD (empirical mode decomposition) and kurtosis rule. Its first step is to reduce the signal noise by the wavelet packet, and then do EMD decomposition. Based on the characteristic that the kurtosis is very sensitive to impact the biggest IMF component of kurtosis is selected to do Hilbert envelope demodulation. As a result, the fault feature information of rolling bearing was obtained. The implementation process of this method was analyzed by simulation signal and the method was successfully applied in inner race of rolling bearing fault diagnosis. © 2013 Springer Science+Business Media New York.