An Asynchronous Induction Motor Fault-Diagnosis Algorithm Based on EMA-2DMSCNN-SENet

With the rapid development of big data and artificial intelligence, motor fault diagnosis has become increasingly important in the fields of intelligent transportation and new energy vehicles. As a core component of the electric vehicle drive system, the operating condition of an asynchronous AC motor is directly related to the safety and reliability of the vehicle powertrain. Once a motor fault occurs, it may lead to powertrain failure, thereby causing traffic accidents, financial losses, and even threats to human life, particularly under high-speed driving conditions where the safety risks are more severe. Therefore, the timely and accurate diagnosis of motor faults in electric vehicles, especially autonomous vehicles, is of great practical significance. To effectively capture the fault-related features embedded in the vibration and voltage signals of asynchronous AC motors and efficiently perform fault diagnosis, this paper introduces a fault-diagnosis model that integrates the exponential moving average (EMA), a two-dimensional multi-scale convolutional neural network (2DMSCNN), and the Squeeze-and-Excitation Network (SENet) mechanism. Experimental validation based on a publicly available asynchronous AC motor fault-diagnosis dataset demonstrates that, compared with traditional machine learning models and ensemble learning methods, the proposed EMA-2DMSCNN-SENet model achieves higher diagnostic accuracy and stronger robustness.