Local Mean Decomposition and Weighted Kurtosis Index for Bearing Defect Detection

Abstract

Vibration signal analysis is an effective technique for detecting bearing defects, and it proposes an approach that involves signal processing methods to extract information about the defects. The initial step in the proposed approach involves dividing the signal into several components (PF) using the LMD algorithm. Subsequently, the weighted kurtosis index (WKI) values are computed, and the summation of components having WKI values higher than the average of WKI leads to the formation of a new signal containing multiple pulses that are very similar to the original signal. Then we can observe peaks at the frequencies of the bearing defects in the envelope spectrum of the new signal. Applying the proposed approach to the vibration signal available in the XJTU database shows a peak at the fault frequency of the outer ring.