The dataset used in this paper contains a total of two , one is the bearing dataset from the University of Paderborn, which uses a total of five KA01,KA04,KA05,K003,KI01 and the author's laboratory self-built dataset, which is divided into a total of six, normal signals, turn-to-turn short circuits, rotor misalignments, rotor breaks, in-bearing damages, and out-of-bearing damages, and both of them use vibration signals as well as current signals were used to generate training as well as experiments.