Ridge-based general synchrosqueezing transform for flexible thin-wall bearing fault diagnosis.

Abstract

Accurate fault diagnosis in flexible thin-wall bearings is crucial for harmonic drive reliability but remains challenging, as fault impulses are often masked by strong operational vibrations. In response to this challenge, a ridge-based general synchrosqueezing transform (RGST) is proposed in this paper. This method unifies time-frequency analysis by operating at the ridge level, using energy trajectories extracted from both instantaneous frequency and group delay estimators. Key features of RGST include a binary ridge expansion mask to enhance energy concentration and suppress noise, and an agglomerative clustering algorithm to separate signal components. Experimental results demonstrate that RGST achieves a concentrated time-frequency representation with superior component separation and noise robustness, thereby improving the reliability of fault diagnosis under multiple fault conditions in flexible thin-wall bearings.