Thermodynamic and environmental analysis of a modern unmixed turbofan engine: application of machine learning

Abstract

This study presents a comprehensive thermodynamic and environmental analysis of modern unmixed turbofan engines, focusing on the application of machine learning to the Pratt & Whitney PW1124GJM engine. High bypass ratio turbofan engines, widely adopted for modern airliners, are designed to optimize fuel consumption efficiency by achieving lower thrust-specific fuel consumption and higher propulsive efficiency compared to traditional turbojet engines. This research investigates the thermodynamic performance and environmental impact of an advanced turbofan engine with unmixed flows. Utilizing GasTurb 13 software, the thermodynamic cycles design performance of these engine are simulated and analyzed. The study further compares flight performance metrics through a hypothetical test scenario, highlighting the potential of machine learning techniques to enhance engine performance evaluation. Machine learning enhances the analysis of a modern unmixed turbofan engine, like the PW1124G-JM, by efficiently processing simulation data to identify performance. This research compares eight machine learning models for predicting turbofan engine parameters (Net Thrust, Specific Fuel Consumption, Fuel Flow, Specific NOx Emissions). Tree-based ensembles (Extra Trees, Gradient Boosting, XGBoost, Random Forest) consistently outperformed other models across all metrics, with Extra Trees and Gradient Boosting achieving perfect R2 scores of 1.0. Support Vector Regression showed moderate performance, while Neural Network results varied significantly between target variables. LightGBM unexpectedly performed poorly despite being tree-based. The findings suggest that for structured engineering datasets with clear physical relationships, tree-based ensemble methods are optimal choices, though the perfect scores warrant further testing for generalizability.