A Method for Predicting Remaining Useful Life of Rolling Bearings Based on Federated Learning Pre-training

Due to the complex operating conditions of rolling bearings and the scarcity of data samples, it is difficult to accurately model the Remaining Useful Life (RUL) prediction models. Additionally, the use of traditional centralized data-driven methods poses data privacy and security issues. Therefore, this paper proposes a method for predicting the Remaining Useful Life of rolling bearings based on federated learning pretraining. Under the distribution of data from different operating conditions at each client, the federated learning concept is used to obtain a fusion model through multiple rounds of model information exchange between clients and the central server. This fusion model serves as the initial model for each client, which then improves the predictive performance for the current bearing by leveraging the degradation information of bearings under other operating conditions contained in the model. Experiments based on the PHM2012 bearing dataset and compared with centralized, local training, and classical federated learning methods show that the method proposed in this paper is effective for bearing RUL prediction.