This study proposes a SCADA-integrated predictive maintenance framework for aircraft engine health monitoring by leveraging machine learning and deep learning models. The objective is to enhance early fault detection capability, reduce unexpected failures, and support cost-efficient maintenance strategies within aviation operations.
A synthetic but operationally realistic time-series dataset was generated using temperature, vibration and fan-speed parameters that are commonly monitored in turbofan engines via SCADA systems. Four algorithms – Support Vector Machine (SVM), Decision Tree (DT), Artificial Neural Network (ANN) and Long Short-Term Memory (LSTM) – were trained and evaluated using accuracy, F1-score, confusion matrices and ROC–AUC metrics. A scenario-based economic assessment was also performed to quantify the financial impact of early fault detection.
Among the evaluated models, LSTM achieved the best overall performance (F1-score = 0.7904; AUC = 0.9834), demonstrating strong capability in capturing temporal degradation patterns. SVM delivered similarly high performance with lower computational load, whereas ANN achieved moderate recall and DT performed the weakest. The economic analysis indicates potential cost savings exceeding US$14m when predictive maintenance actions are applied proactively based on LSTM predictions.
To the best of the authors’ knowledge, this study provides one of the first holistic assessments that combines SCADA-based synthetic data generation, comparative AI model evaluation, and an explicit economic analysis tailored for aircraft engine maintenance. The proposed framework offers a transferable foundation for integrating AI-driven predictive maintenance solutions into real-world aviation environments.
