TMSST-CK fault feature extraction method for flexible thin-wall bearing based on the Gini index principle

Abstract

A flexible thin-wall bearing (FTB) is forced to deform with the elastic deformation of the flexspline in motion, resulting in a completely different fault characteristic frequency (FCF) from ordinary rolling bearings. However, current extraction methods treat FTBs as ordinary ones and they are analyzed based on the spectrum, without considering the time-varying nature of the FCF in an FTB. In an effort to address this concern, this paper proposes a Gini index-based time-reassigned multisynchrosqueezing transform, where the window size is determined by the period impulses feature in the noise-removed signal. In addition, a frequency point selection method that contains fault characteristics of the FTB is proposed by calculating the maximum correlated kurtosis in the time–frequency (TF) plane. Furthermore, the instantaneous frequency is extracted by calculating the time interval between two consecutive impulses. The extracted fault frequency agrees well with the experimental result and outperforms other TF analysis techniques in demonstrating its time-varying nature.