Machine learning-driven optimization of performance and emissions in a butanol/diesel CI engine

A hybrid modeling framework integrating experiments, CONVERGE CFD, and Optuna-optimized ensemble machine learning is developed to predict and optimize performance and emission characteristics of a DI-CI engine fueled with butanol/diesel blends. Machine learning was employed to model and predict key combustion and emission characteristics under varying engine operating conditions. This work presents a numerical and modeling examination of a DI-CI engine's performance and emission characteristics. CONVERGE CFD software was used to do the numerical analysis. Four distinct engine parameters were examined in this study: Start of Injection (SOI) (17 to 29° CA bTDC), Compression Ratio (CR) (14.5 to 19.5), Exhaust Gas Recirculation (EGR) (0 to 30%), and Fuel Injection Pressure (FIP) (200 to 280 bar). The Optuna-optimized ensemble Meta Model achieved the most accurate predictions, exhibiting the lowest RMSE and highest R² values for ISFC and NOₓ, along with reasonable accuracy for Soot estimation. The ensemble Meta Model achieved the best performance with R² values of 0.98 (ISFC) and 0.97 (NOx) and RMSE below 5%, demonstrating strong generalization.