Journal of Machine Learning and Applications

Asaph Matheus Barbosa1*, Thao Vy Nhat Ngo1, Elaheh Jafarigol1, Theodore B. Trafalis1, Emuobosa P. Ojoboh1

The aim of this research is to predict the Remaining Useful Life (RUL) of turbine jet engines using a federated machine learning framework, ensuring data privacy and security while maintaining high predictive accuracy. Federated Learning (FL) enables multiple edge devices/nodes or servers to collaboratively train a shared model without sharing sensitive data, making it ideal for industries like aviation, where data confidentiality is paramount. The proposed system employs Long Short-Term Memory (LSTM) networks, a type of recurrent neural network, to model the complex temporal relationships and degradation patterns in engine data. By leveraging decentralized computation, the framework allows models to be trained locally on each device, with learned weights aggregated at a central server using the Federated Averaging (FedAvg) algorithm. The study utilizes the C-MAPSS (Commercial Modular Aero-Propulsion System Simulation) dataset, a publicly available resource from NASA, which simulates engine degradation under various operational conditions. The dataset includes time-series data from 21 sensors and three operational settings, providing a comprehensive foundation for analyzing fault progression and failure modes. Computational results demonstrate the effectiveness of the proposed approach; with validation and test Root Mean Squared Error (RMSE) metrics range from 13.811 to 22.998 across different operational scenarios. The aggregated FL model achieved an RMSE of 17.899, showcasing