<ul><li>Aircraft engines are continuously monitored using sensors to predict maintenance needs and Remaining Useful Life (RUL). Limited run-to-failure data samples challenge the development of prognostics.</li><li>A collaborative federated learning framework is proposed where multiple airlines cooperate to train a collective RUL prognostic machine learning model without sharing data centrally due to privacy concerns.</li><li>Four novel methods are introduced to aggregate the parameters of the global prognostic model, enhancing its robustness against noisy sensor data.</li><li>The framework is tested using the N-CMAPSS dataset with six airlines collaborating in the federated learning approach, resulting in more accurate RUL prognostics compared to individual airline models and robustness to noisy data samples.</li></ul>

Federated learning framework for collaborative remaining useful life prognostics: an aircraft engine case study

Discover more