Fault Detection of Permanent Magnet Synchronous Motor Based on Deep Learning Method

This paper proposes a deep learning algorithm for motor fault detection. Based on the Long Short-Term Memory (LSTM), one of the deep learning algorithm, catching the three-phase current values and the information of electrical angle in the previous sampling instants, the three-phase current value at the next sampling instant can be predicted in real time; the predicted error is not affected by torque fluctuations. Therefore, the operating status of the motor can be observed in real time. The simulation results show that the error waveform amplitude is very small when the motor is running normally with the load torque randomly fluctuating, and the error amplitude will increase sharply when the motor is going to be broken down.