Machine-learning-assisted optimization of alumina-enhanced Citrullus lanatus seed biodiesel combustion in a thermal-barrier-coated diesel engine

The performance and emission characteristics of a thermal barrier-coated CI engine operating with watermelon seed oil methyl ester (WSOME) biodiesel blends with Al2O3 nanoparticles were investigated. The various biodiesel blends (B10, B20, and B30) were used with different concentrations of Al2O3 (25 ppm, 50 ppm, and 75 ppm) at different engine loads. Response Surface Methodology (RSM) has been used to identify optimal conditions to maximize brake thermal efficiency (BTE), minimize specific fuel consumption (SFC), and reduce CO, HC, NOx, and smoke emissions. The best condition was identified as B30 (75 ppm Al2O3) at 50 % load, with a desirability value of 0.757 based on RSM optimization results. Under these conditions, the predicted responses were SFC of 0.317 kg/kWh, BTE of 25.063 %, CO of 0.0219 vol%, HC of 30.75 ppm, NOx of 408.46 ppm, and smoke of 27.08 HSU. Optimized conditions yielded higher BTE and lower SFC, CO, HC, and NOx emissions than the baseline (uncoated diesel). Smoke increased slightly. Finally, an ANN–XGBoost stacking model was developed as a predictive modelling approach to estimate performance and emission responses as functions of blend ratio, engine load, and nanoparticle concentration. The ML results also agreed with the experimental and RSM trends, revealing that combustion in a TBC engine with Al2O3 can effectively enhance fuel economy and reduce major gaseous emissions within the engine’s tested range.