MAFL-PdM: A Multi-Agent Federated Learning Framework for Personalised Predictive Maintenance Across Heterogeneous Industrial IoT Sites
Abstract
Abstract Industrial Internet of Things (IIoT) environments increasingly rely on predictive maintenance (PdM) to reduce unplanned downtime and improve operational efficiency, yet cross-site collaboration remains constrained by data ownership and heterogeneity. This paper proposes MAFL-PdM, a multi-agent federated learning framework for personalised predictive maintenance across heterogeneous industrial IoT sites. Four autonomous Factory Agents and a Federated Orchestrator Agent collaboratively estimate remaining useful life (RUL) without transmitting raw sensor data, combining a two-layer long short-term memory network with temporal attention, weighted Huber loss, FedProx regularisation, a clustering-driven aggregation mechanism (CMFL-PdM), drift-aware client selection, and a post-federation personalisation step. Experiments use two heterogeneous real-world industrial datasets - elevator door sensor data and metalworking-fluid process monitoring - under stratified random 80/20 splits, with all results reported as means over five random seeds. MAFL-PdM achieves an RMSE of 0.0498 ± 0.0012, a mean absolute error of 0.0348 ± 0.0008, an F1 score of 0.819 ± 0.016, and a recall of 0.773 ± 0.012, against a naive mean-predictor RMSE of 0.248. It attains a lower RMSE than centralised training in all five seeds while matching it on F1, and outperforms every federated baseline by a wide margin. A component ablation at the same five seeds localises the effect precisely: post-federation…
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.