Intelligent Fault Diagnosis of Bearing Based on Gramian Angular Field Image and Improved MobileViT

Bearings are crucial components in industrial equipment, and their health monitoring and fault diagnosis are essential for ensuring system stability and safety. Moreover, feature extraction is a key step affecting the diagnostic performance of the model. To improve fault diagnosis accuracy, this paper proposes a bearing fault diagnosis method based on image information and an improved MobileViT network. The method converts signals from multiple sensors into Gramian angular difference field (GADF) images, which are then processed by a MobileViT model with an enhanced attention mechanism (AAM) for efficient fault diagnosis. Experimental results using the Southeast University bearing dataset, compared with CNN and MobileViT, as well as two other image transformations, i.e. scatter plot (STD) and relative position matrix (RPM), demonstrate that the proposed method significantly improves diagnosis accuracy.