Improved Fault Diagnosis Method for Permanent Magnet Synchronous Machine System Based on Lightweight Multisource Information Data Layer Fusion
Jun Hang, Gaopeng Qiu, Menglu Hao, Shichuan Ding
Abstract
Fault diagnosis is essential for the safe operation of a permanent magnet synchronous machine (PMSM) system. At present, the fault diagnosis method based on deep learning has been gradually studied due to its ability to automatically extract fault features and an end-to-end diagnostic model. However, this method often uses a single-source signal, resulting in poor fault diagnosis performance. On the other hand, deep learning model typically requires large amounts of storage and computational resources, especially when processing multisource signals. Hence, this article proposes an improved fault diagnosis method for PMSM systems based on lightweight multisource information data layer fusion. In this method, the original multisource information is fused at the data layer. Then the fused information is input into the constructed lightweight deep learning model to implement fault diagnosis of PMSM system, where a one-dimensional convolutional neural network with the combination of depth-separable convolution and global average pooling is first presented to reduce the computational time and complexity of the fault diagnosis method. The simulation and experimental results show that the proposed fault diagnosis method can achieve accurate fault identification of the PMSM system and can reduce the requirement for storage and computational resources, indicating the effectiveness of the proposed fault diagnosis method.