A Novel Reinforcement Learning via Dual-Engine Quick Method for Motor Bearings Unknown Faults

As one of the most core and vulnerable components of electric vehicles (EVs), the failure of bearings has become a key issue limiting the safety of vehicle systems. Quick diagnosis of faults can ensure the continuous and stable operation of EVs. However, the complexity of the driving environment and the dynamic nature of EVs lead to unknown bearing failures. This seriously affects the performance of the vehicle system and personal safety. In response to the above issues, a dual-engine quick diagnosis method for unknown faults in motor bearings via the chord transform rotary evaporation is proposed. First, the rotary evaporation strategy is proposed. This method meets the requirements of reinforcement learning on the amount of unlabeled sample data and effectively solves the problem of confusing old and new samples in the learning process. Second, an adaptive chord transform feature affine-invariant strategy is proposed. This method better accepts the constraints between evaporation loss and bearing state category loss and effectively solves the data computation problem. Finally, a comparison is made with existing advanced methods in a multioperating condition open dataset and an experimental platform for bearing faults. The experimental results validated the advantages of the dual-engine model in diagnosing unknown faults of motor bearings.