Predicting unscheduled aircraft maintenance orders: A comparison study between machine learning techniques

Ensuring aircraft reliability and reducing downtime is essential in flight operations. This study applies and investigates the performances of several machine learning techniques namely regression learner, gradient boosting and artificial neural networks (ANNs) to predict unscheduled maintenance orders for a leading airline company in Saudi Arabia. In this study, we aim to develop a robust and accurate artificial intelligence (AI)-based model that predicts unscheduled aircraft maintenance orders that can help operational managers to construct a reliable maintenance plan. The models are trained using historical maintenance data and flight parameters to identify patterns leading to unscheduled maintenance orders. The findings revealed that ANNs yielded a substantial improvement in prediction accuracy compared to regression learner and gradient boosting, which could improve the operational effectiveness, increase safety, increase customer satisfaction, and reduce operational costs. All employed models are evaluated on four aircraft: A320, A333, A777, and A787. The models' errors are compared with their mean absolute deviation (MAD), mean square error (MSE), and mean absolute percentage error (MAPE). The results show that the ANN has the lowest MSE in all aircraft types, showing that the ANN is a more accurate technique in predicting unscheduled aircraft maintenance orders.