Fault diagnosis of automobile drive based on a novel deep neural network

Abstract

Abstract The times are progressing. Facing the increasing number of electric vehicles, they use power batteries as energy storage power sources. As a core component of electric vehicle, the drive motor is related to the normal operation of the vehicle. If the driving motor fails, passengers may be irreversibly hurt, so it is very important to diagnose the driving motor of electric vehicle. This paper mainly analyzes the faults of electric vehicles, and makes use of diagnostic signals to diagnose the faults. A novel fault diagnosis method of automobile drive based on deep neural network is proposed. In this method, CNN-LSTM model is constructed. Firstly, the vibration signals are transformed into time-frequency images by fast Fourier transform, and then the time-frequency images are input into the proposed model to obtain the fault classification results. In addition, CNN, LSTM and BP neural network are introduced to compare with the methods proposed in this paper. The results show that CNN-LSTM model is superior to the other three models in the fault diagnosis of automobile drive, reaching 99.02 % of the fault accuracy rate, showing excellent fault diagnosis performance. And when the same learning rate is used for training, the rate of loss reduction is obviously better than that of the other three types of vehicle drive fault diagnosis method based on CNN-LSTM.