Performance and Emission Prediction of CI Engine Fuelled with ZOME20 Blends Enhanced by Nanoparticles using Advanced Machine Learning Models

Abstract

Here, we investigate the joint effects of biodiesel-nanoparticle fuel composites and machine learning modelling on diesel engine performance and emissions. The experiment was conducted using a single-cylinder diesel engine with 20% Ziziphus Oenoplia Methyl Ester (ZOME20) and is paired with 80% diesel. Al₂O₃ and graphene oxide nanoparticles were added to the ZOME20 blend at the concentration levels of 50-200 ppm. As the results show, due to a high thermal conductivity and accompanying catalytic oxidation capacity, Al₂O₃ nanoparticles significantly improved combustion efficiency. Meanwhile, 20% Ziziphus Oenoplia Biodiesel Blend with 150 ppm Additive (ZO20A150) yielded the highest Brake Thermal Efficiency (BTE) and the lowest Brake Specific Fuel Consumption (BSFC); it also helped decrease CO, HC and smoke emissions, while the tendency of NOx also decreased, which was attributed to the in-cylinder temperature rise. Opposite results could be obtained in the models, with KNN, Hyper-SVR and linear models delivering the most credible results and undercutting the experimental data by 10% at maximum. The best prediction was made by the Random Forest model, with R² ≈ 1.0. Thus, the investigation is indicative of ZO20A150 being the best blend with which to improve efficiency at the expense of increased emissions and the Random Forest model is the most reliable ML tool for measuring engine performance when biodiesel is matched with specific types of nanoparticles.