Train axlebox bearing composite fault denoising method based on SVMD and dual threshold classification criteria

Abstract The increase in train operating speed and the alteration of the wheel-rail contact relationship, which have resulted in a deterioration of the service conditions of axlebox bearings, render the composite bearing failure characteristics susceptible to being obliterated by robust background noise. In order to effectively filter out redundant noise and accurately extract the fault characteristics, a dual-threshold signal denoising method is constructed based on successive variational mode decomposition (SVMD) and combining the Euclidean distance and kurtosis characteristic (Dual-EDK). The validity of the proposed method is effective by the composite fault data of axlebox bearings on the equal scale test bench. Firstly, the signal is adaptively decomposed into K components by SVMD, which are classified into three categories by Dual-EDK: effective IMF, noisy IMF and noise IMF. Then, the effective IMF is retained, while the noisy IMF is discarded. Wavelet soft threshold (WST) is employed to reduce the noisy IMF, reconstruct effective IMF and denoising noisy IMF, and envelope demodulation is used to achieve noise reduction. Finally, the composite fault signals at both normal operating speeds and critical speed of instability are collated and subjected to analysis by the test bench. This is done in order to ascertain the efficacy of the proposed method by comparing the results with those obtained through the traditional approach. The results of the analysis demonstrate that, in comparison to VMD, the SNR is enhanced by 27.6% and the
RMSE is diminished by 7.4% at normal operating speeds, the SNR is augmented by 26.2% and the RMSE is reduced by 9.6% at the unstable critical speed. These outcomes illustrate that the proposed method is capable of effectively extracting the denoising characteristic components of bearing faults, and that it has a pronounced impact.