Exploitation of a Multiband Weighted Envelope Spectrum for Railway Axle Bearing Fault Diagnosis

Accurate diagnosis of axle bearing faults is a key part of ensuring the operating safety of railway rolling stock. The methodology combining spectral coherence and envelope analysis proved to be efficient for detecting bearing defects. However, the fault-related information of the axle bearing is disturbed by the complex noise in the train and exhibits a multi-band distribution, which cannot be revealed through the traditional envelope spectrum based on spectral coherence. To overcome these limitations, a spectral coherence-based multiband weighted envelope spectrum (MWES) is presented for the detection of axle bearing defects, enabling efficient extraction and integration of information in multiple frequency bands. Firstly, a signal-to-noise ratio defined in the frequency domain (FDSNR) is introduced for evaluating the richness of fault-related information along the spectral frequency in the spectral coherence. Secondly, a weighting function is designed for the FDSNR using a threshold denoising strategy that enhances the fault-related information while removing the interfering components. Finally, MWES for axle bearing diagnosis is constructed by weighted integration of the normalized spectral coherence along spectral frequency. The fault diagnosis performance of MWES is validated by experimental data from axle bearings. Comparisons with state-of-the-art envelope spectrum techniques highlight the advances of the developed MWES.