Adaptive filtering enhanced windowed correlated kurtosis for multiple faults diagnosis of locomotive bearings

Compound faults diagnosis of locomotive bearings are still a challenge especially when the multi-fault impulses share the common resonant frequency. In this paper, an adaptive filtering enhanced windowed correlated kurtosis (WCK) method is proposed to isolate and identify each fault mode. A concept termed flexible analytical wavelet transform (FAWT) spectrum is defined to construct proper FAWT basis such that high signal-to-noise (SNR) is obtained in the filtered signal. Further, WCK is applied on the denoised signals to successively isolate each fault mode and determine the defects number. The performance of the proposed method is validated via analyzing experiment measurements from the locomotive bearings subjected to two local roller defects and three local outer race damages.