Abstract This study presents a federated learning (FL) framework for predictive maintenance (PdM) of satellite battery systems, integrating the NASA BP930 dataset with MACCOR-generated simulations to create a hybrid training environment that reflects both real and synthetic operating conditions. Unlike previous FL studies in PdM, the proposed approach models resource constraints such as power variability and communication limits, replicating conditions in large-scale industrial and aerospace networks. Feedforward neural networks (FFNNs) are deployed at edge nodes, and global aggregation is performed through a reliability-weighted FedAvg algorithm. Experimental results show that our proposed framework achieves 95.0% predictive accuracy under ideal conditions and achieves 90.1% training convergence accuracy in constrained networks, reflecting exceptional robustness of the learning process while maintaining reliable predictive test performance. Comparable studies on federated battery prediction generally report accuracies within the 93–97% range under controlled laboratory settings, without assessing power or communication limitations. The proposed framework therefore attains equal or higher predictive precision under ideal conditions, while maintaining stable convergence under degraded environments. It records an F1-score of 0.92, outperforming standard local RNN and LSTM baselines by 6–8% and ensuring complete data privacy. The novelty of this work lies in its realistic simulation of industrial deployment scenarios, hybrid data integration, and systematic testing under constrained environments, all of which are not sufficiently addressed in existing FL-based PdM research.
Adel et al. (Wed,) studied this question.