A Federated Learning-Based System for Electric Vehicle Battery Health Prediction and Fault Diagnosis
Year: 2026
Type: conference-paper
Abstract: The management of the health of industrial equipment faces the dual challenges of protecting data privacy and enabling multi-site collaborative modelling. This paper proposes an architecture for an equipment health management system based on federated learning. This system achieves distributed collaborative modelling by carrying out local training at edge nodes and aggregating parameters at a central server. A health status prediction model is constructed by the system using an LSTM network, which captures equipment degradation trends. A convolutional neural network is used for fault pattern classification and diagnosis, and a differential privacy mechanism is employed to ensure the security of model training. It has been demonstrated by experiments that this approach effectively reduces communication overhead while prediction accuracy is maintained, issues associated with data privacy leakage are addressed that are associated with traditional centralized training, and a new approach for the collaborative utilisation of crossenterprise equipment health data is provided.
Institution: Jiangmen Polytechnic
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7204090752