Machine Learning–Based Prediction and Interpretability Analysis of Aircraft Engine Landing and Take-Off Emissions

This study presents a comprehensive data-driven framework for predicting aircraft engine emissions under Landing and Take-Off (LTO) conditions using certified data from the ICAO (The International Civil Aviation Organization) Aircraft Engine Emissions Databank. The primary objective is to develop reliable machine learning (ML) models for estimating emission indices of nitrogen oxides (NOx), carbon monoxide (CO), and unburned hydrocarbons (HC), while simultaneously providing physical interpretability of model behaviour. The methodology involves systematic data preprocessing, feature selection, and the application of multiple ML algorithms, including Random Forest, Gradient Boosting, Support Vector Machine, Artificial Neural Network, and Linear Regression models. Model performance was evaluated using statistical metrics and Taylor diagram analysis, while interpretability was assessed through feature importance, partial dependence plots, individual conditional expectation plots, and error variability analysis across thrust classes and LTO phases. Quantitative results show that ensemble models outperform other approaches, with the Random Forest model achieving correlation coefficients of approximately 0.95 for NOx, 0.80 for CO, and 0.70 for HC, along with superior variance representation. Fuel flow emerged as the most influential parameter, contributing nearly 80% for NOx, over 60% for CO, and approximately 45% for HC, while thrust and pressure ratio exhibited pollutant-specific secondary influences. Error variability analysis revealed higher uncertainty for CO and HC during idle and low-thrust regimes, whereas NOx prediction uncertainty was dominated by climb-out and take-off phases. The study concludes that data-driven ensemble ML models, combined with interpretability analysis, provide a robust and physically consistent approach for aircraft engine emissions prediction, offering valuable insights for emission mitigation strategies and regulatory assessment.