Development of Deep Learning-Based Cooperative Fault Diagnosis Method for Multi-PMSM Drive System in 4WID-EVs

A deep learning-based cooperative fault diagnosis method is proposed in this study. The proposed method aims to improve the reliability of contemporary fault diagnosis methods for the multipermanent magnet synchronous motor (PMSM) drive system of four-wheel-independent-drive electric vehicles (4WID-EVs). Specifically, a novel 1-D signal to 2-D RGB image conversion method is proposed to transform the fault diagnosis task into an image classification problem, and the impressive image classification capability of the convolutional neural network (CNN) is utilized to achieve a highly accurate fault diagnosis function. Following this, CNNs of various architectures are trained using the constructed RGB image dataset, and the fault diagnosis accuracy under different architectures is compared. In addition, the adaptability of the trained network is increased by utilizing two adaptive factors to ensure that the trained CNN can be compatible with all untrained combinations of angular speed and load torque. Furthermore, by combining the designed single fault diagnosis method with directed-graph theory, a comprehensive evaluation criteria-based cooperative diagnosis method is proposed, which further improves the accuracy significantly. The effectiveness of the proposed fault diagnosis method is demonstrated via adequate comparison.