Fault Diagnosis of Bearing Based on Multi-Channel Image Fusion and Deep Learning

In the task of machinery fault diagnosis, single-channel signal often fails to comprehensively reflect the operational state characteristics of the equipment. To improve diagnostic accuracy and more precisely represent the equipment health status, this paper proposes a bearing fault diagnosis method based on multi-channel image fusion and GADF-MobileVit. The method first converts raw signals collected by multiple sensors into gramian angular difference field (GADF) images and then fuses the GADF images from different channels. Finally, the fused GADF images are fed into the MobileVit network model to achieve efficient bearing fault diagnosis. Validation on the Southeast University bearing dataset, along with comparisons between single-channel data and multi-channel data fusion results, demonstrate that multi-channel data fusion significantly enhances the model's performance in bearing fault diagnosis.