Federated Learning for Rolling Bearing Fault Diagnosis: A RepViT-Enabled Approach

Federated learning (FL) has emerged as a promising approach for fault diagnosis in rotating machinery, addressing data privacy and availability challenges. This paper presents a federated learning-based fault diagnosis framework, leveraging lightweight Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) to diagnose faults under varying operating conditions accurately. The proposed methodology utilizes the RepViT model, a hybrid architecture combining MobileNetV3_L and lightweight ViTs, to achieve efficient fault diagnosis on mobile devices. Experiments conducted on the HUST-bearing dataset demonstrate the effectiveness of the proposed method, achieving high diagnostic accuracy even with limited data. The study further explores the impact of varying the number of participating clients and the robustness of the model under different client availability scenarios. Results show that the federated approach maintains high diagnostic performance while preserving data privacy, making it a viable solution for industrial applications.