Hybrid XGBoost–ANN modelling and optimization of TiO2-enhanced Mahua biodiesel for improved engine performance and emissions

The growing demand for cleaner transportation fuels and the depletion of fossil resources have intensified global interest in renewable biodiesel. This study investigates the performance and emission characteristics of a compression ignition (CI) engine fuelled with Mahua Oil Methyl Ester (MOME) blended with titanium dioxide (TiO 2 ) nanoparticles. Experiments were conducted on a single-cylinder, four-stroke CI engine using B10, B20, B30, and B40 blends under varying load conditions. Engine responses were evaluated in terms of brake thermal efficiency (BTE), brake specific fuel consumption (BSFC), exhaust gas temperature (EGT), and major pollutants (NOx, CO, HC, and smoke). While the combustion performance of the TiO 2 -enhanced biodiesel has been proven to be promising, experimental assessment of the engine performance and the hybrid machine learning prediction, as well as multi-objective optimization, in a single framework has been limited. Under full-load conditions, the brake thermal efficiency of B30 – TiO 2 was found to be around 38.0% which is almost 8.6% higher than neat diesel among the fuel blends investigated, and it was also observed that it reduced the brake specific fuel consumption and exhaust emissions. The proposed model (Hybrid XGBoost–ANN) showed excellent predictive accuracy with R 2 = 0.971, RMSE = 28.7 ppm, MAE = 22.3 ppm, SEP = 24.1 ppm and AAD = 2.84%, which was better than the individual XGBoost and ANN models. The optimized machine learning model was combined with NSGA-II to determine the Pareto-optimal operating parameters and the optimum operating parameters was determined by using desirability function, while Response Surface Methodology was used to visualize the interactions among operating variables at a 28% blend ratio and 80% engine load, with a composite desirability score of 0.94. The engine performance and emission performance of the blends containing TiO 2 are promising compared to the conventional diesel fuel under the studied operating conditions as concluded from the combined experimental, machine learning and optimization analysis.

Перевод пока недоступен