Multisensor-Driven Intelligent Mechanical Fault Diagnosis Based on Convolutional Neural Network and Transformer

Deep learning (DL) has been widely used for intelligent fault diagnosis of rotating machinery. Nevertheless, the diagnosis performance is usually impacted by varying working conditions and noise interference. To address this issue, this article proposes a multisensor-driven intelligent fault diagnosis method based on convolutional neural network (CNN) and Transformer. Specifically, a signal-to-image conversion method based on truncated singular value decomposition (SVD) and Gramian angular field (GAF) is constructed to fuse multisensor time-series signals into color images. By the building and integration of a convolution embedding unit and a lightweight Transformer encoder (LFormer encoder), a lightweight convolutional Transformer for feature extraction and classification is established, which could efficiently learn both local and global features from the color images. Experimental studies are conducted on two fault diagnosis datasets to verify the effectiveness and superiority of the proposed method, and the results demonstrate that the proposed method is superior to the existing methods, especially under varying working conditions and noise interference.