Time–frequency-ridges-based synchrosqueezing transform for flexible thin-wall bearing fault diagnosis

Abstract

Within the realm of fault diagnosis, the flexible thin-wall bearing, with its unique kinematic attributes diverging from those of rolling bearings, presents a challenge. The reassignment method, as a post-processing technique in time-frequency analysis, enhances time-frequency representation readability but compromises signal reconstruction. To address this, a time–frequency-ridges-based synchrosqueezing transform is proposed in this paper. Initially, to identify periodic components, the estimation of instantaneous frequency and group delay is obtain as time–frequency ridges from wavelet transform. Then, a localization method is employed to lower the error between the estimation ridges and the true ridges. Furthermore, an energy-intersection method is designed to divide the ridges into impulse and harmonic types. On this basis, the proposed method reassigns the energy of impulsive components and harmonic components separately. Consequently, an improvement in the readability of time-frequency representation is achieved, alongside the preservation of reconstructive ability. Experimental validation using simulated and vibration signals underscores its efficacy in identifying periodic components and providing precise diagnostics for flexible thin-wall bearings.