Abstract The increasing focus on the United Nations Sustainable Development Goals and on global net-zero emission targets has intensified the search for renewable and low-carbon alternatives to fossil diesel in compression ignition engines. Biodiesel made from non-edible feedstocks is an acceptable route to lowering greenhouse gas emissions and compatibility with existing engine infrastructure. In the present study, the engine performance and emission characteristics of biodiesel produced from Mesua ferrea seed oil were experimentally studied. The biodiesel was characterized in terms of some of its main physicochemical properties and tested in a single-cylinder diesel engine under various load conditions and fuel injection pressures of 200–240 bar. Engine performance parameters such as brake thermal efficiency and brake specific fuel consumption were studied in addition to exhaust emissions such as CO, HC, and NOx. The results show that the highest pressure of injection improves the efficiency of combustion and thus thermal efficiency and fuel consumption of the brake. Carbon monoxide and hydrocarbon emissions were greatly reduced with increasing injection pressures, while NOx showed an increasing trend with increasing in-cylinder temperatures. This study applies explainable machine learning to predict engine performance and emissions. XGBoost outperformed Linear Regression and Decision Tree (R²ₜest: 0. 932–0. 969), with comprehensive XAI validation. Partial Dependence Plots revealed Load’s linear BTE/BSFC/NOx control and FIP’s non-linear CO/UHC suppression. SHAP analysis quantified Load’s 82% average attribution versus FIP’s 18%, corroborated by permutation importance. PDP convexity confirmed NOx thermal kinetics and CO oxidation limits.
Paramasivam et al. (Wed,) studied this question.