A method and application for fault diagnosis of coal mine machinery bearings under nonstationary state conditions

YAN Zhen; WANG Yuanzhi; GENG Pei'en; YU Gang

doi:10.16578/j.issn.1004.2539.2026.06.019

Abstract：[Objective]Aiming at the time-frequency energy diffusion phenomenon induced by the limitation imposed by the Heisenberg uncertainty principle in traditional wavelet transform,a high-resolution time-frequency analysis framework was constructed to break through the energy divergence problem of existing methods for strongly frequency-varying impact signals.[Methods]Firstly,a transient wavelet transform algorithm was proposed to improve the time-frequency aggregation of wavelet transform and enhance its analytical capability for impact signals.Secondly,a second-order group delay estimation operator was calculated via wavelet transform to characterize transient events in impact signals.Then,a fixed-point iteration strategy was adopted to enhance the estimation accuracy of this operator for strongly frequency-varying signals.Finally,a time-direction compression strategy was integrated to provide high-resolution time-frequency representations for strongly frequency-varying impact components in signals.[Results]The results show that the proposed method maintains favorable performance under strong noise and severe group delay variation conditions.The Rényi entropy is reduced by approximately 34.0%compared with the wavelet transform-based time-reassigned synchrosqueezing transform,and the convergence speed of normalized energy is increased by more than 40%.In the constant-speed bearing fault test,the 9.4 ms fault characteristic interval can be accurately extracted.Under variable-speed conditions,the dynamic fault characteristic frequencies of 137.93 Hz and 106.40 Hz can be precisely identified,and its running time is only about 37.5%of that of 3rd-order synchroextracting transform,which provides a reference for transient signal analysis in mechanical fault diagnosis.