“…However, the working environment of most bearings is complex, and the collected signals contain a lot of noise and harmonic components, which cause obstacles for Processes 2022, 10, 1734 2 of 25 the accurate extraction of bearing faults [3]. Many methods have been applied to improve the accuracy of fault diagnosis and remaining life prediction in complex environments, such as dynamic model analysis based on the physical characteristics of the bearing itself [4][5][6], signal analysis methods in the time-frequency domain [7][8][9], methods based on entropy of the information contained in the signal [10,11], end-to-end methods of the neural network [1,12,13], and the method of model data fusion [14,15].…”