From Preventive to Predictive: Advancing Timely and Accurate Aircraft Prognostics

The aviation industry heavily relies on effective maintenance strategies to ensure operational safety, but traditional preventative maintenance procedures often result in higher economic burdens, increased operational disruptions, and flight cancellations. Predictive maintenance offers a transformative alternative by forecasting equipment failures, optimizing schedules, and enhancing safety. In this study, we propose and evaluate three novel variants of the Gradient-Boosting Regression Tree algorithm for the prediction of aircraft turbofan engines' Remaining Useful Life (RUL). Our research investigates how feature engineering, operating conditions, and fault modes affect predictive performance across diverse engine operating scenarios. We conduct a comprehensive comparative analysis of our proposed algorithms against seven state-of-the-art methods, including traditional machine learning, deep learning, and hybrid architectures. Additionally, we introduce the Margin-Adjusted Reliability Score (MARS), a novel benchmarking metric that incorporates both prediction accuracy and timeliness, addressing gaps in existing evaluation methods. By providing insights into algorithm interpretability and performance, this work contributes to the development of efficient, transparent, and industry-relevant predictive maintenance solutions, advancing the state of fault prognostic systems in aviation. The datasets, tools, and algorithms from this work will be open-sourced to support community research.