A Federated Learning Method with DNN and 1DCNN Feature Fusion for Multiple Working Conditions Fault Diagnosis

Under multiple working conditions, the sample size of each client is small and it is not easy to obtain data from other working conditions, making it difficult to establish an effective deep learning fault diagnosis model. Federated learning is a distributed training method that can accomplish collaborative training of multiple clients without directly sharing client data. This paper proposes a federated learning method with DNN and 1DCNN feature fusion. This method designs a feature fusion network based on 1DCNN and DNN, and establishes a federated learning architecture for the feature fusion network in order to better extract fault features and improve the accuracy of multiple working conditions fault diagnosis. The efficiency of the proposed method is demonstrated by utilizing the Case Western Reserve University bearing data set.