“…A wide range of studies have dedicated on advancing the bearing fault diagnosis capacity, where the machine learning approaches play an important role (Dong et al, 2021; Duan et al, 2021; Zhou et al, 2017). The underlying idea of these approaches is to elucidate the intrinsic correlation between the faults and different types of measurements, such as vibration, acoustic emission and eddy current and so on (Aasi et al, 2021; Ben Ali et al, 2015; Chen et al, 2016; De Moura et al, 2011; Jiang et al, 2019; Pandya et al, 2013; Tabatabaei et al, 2020; ). Among them, vibration signals are most widely used for bearing fault diagnosis because of the low instrumentation cost and sufficient fault-related signatures contained (Ben Ali et al, 2015; Chen et al, 2016; De Moura et al, 2011; Liang and Zhou, 2021).…”