Seasonal Feature Augmentation for Interpretable Turbofan RUL Prediction via Bayesian-Optimized XGBoost

The operational safety and sustainability of gas turbine engines in the aviation industry are directly contingent upon the accurate prediction of the Remaining Useful Life (RUL) of critical components. Existing RUL prediction models in the aviation literature predominantly focus on in-engine sensor data, largely overlooking the cumulative impact of external environmental variables—particularly ambient temperature and seasonal fluctuations—on material fatigue. To address this gap, the present study proposes a novel hybrid model grounded in the NASA Commercial Modular Aero-Propulsion System Simulation (C-MAPSS) dataset (FD001–FD004), integrating standard operational data with seasonal parameters. The proposed model architecture is based on the Extreme Gradient Boosting (XGBoost) algorithm, with hyperparameter optimization conducted via the Bayesian-based Optuna framework. Experimental findings demonstrate that the model achieves RMSE = 15.84 and R² = 0.85, exhibiting competitive or superior predictive accuracy relative to complex deep learning architectures reported in the literature. Furthermore, the study explores the contribution of a synthetically constructed periodic feature to the predictive behavior and interpretability of the model.