Machine Learning Prediction and Optimization of Performance and Emissions Characteristics of IC Engine

In this work, a study was conducted to investigate the effects of different biodiesel blends with hydrogen peroxide additive on performance and emissions an internal combustion engine under various operating parameters. A CI operated diesel, four dissimilar biodiesels, H2O2 at proportions. The used were Jatropha (D60JB30A10, D60JB34A6, D60JB38A2, D60JB40), Honge (D60HB30A10, D60HB34A6, D60HB38A2, D60HB40), Simarouba (D60SB30A10, D60SB34A6, D60SB38A2, D60SB40), Neem (D60NB30A10, D60NB34A6, D60NB38A2, D60NB40). tested injection pressures (200, 205, 210 bar), speed 1500 rpm, CR 17.5:1. From experiments conducted, it highlighted that, specific conditions, i.e., pressure 205 bar, 80% load, compression ratio 17.5, timing set 230 before top dead center, D60JB30A10, D60HB30A10, D60SB30A10, D60NB30A10 achieved highest brake thermal efficiencies 24%, 23.9675%, 23.935%, 23.9025%, respectively. Notably, blend D60JB30A10 stood out efficiency 24% among these blends. Similarly, when evaluating same operational exhibited lowest levels: CO (0.16% Vol), CO2 (7.8% HC (59 PPM), Smoke (60 HSU), while NOx (720 PPM) showed relative increase higher concentrations hydrogen-based additive. in comparison. Additionally, suggests that machine learning techniques can be employed predict emission characteristics, thereby cutting down time costs associated traditional trials. Specifically, methods, like XG Boost, random forest regressor, decision tree linear regression, utilized for prediction purposes. Among techniques, Boost model demonstrated highly accurate predictions, followed by regression models. accuracy predictions assessed through evaluation metrics such as R2-Score (0.999), Root Mean Squared Error (0.540), (0.248), Absolute (0.292), which allowed thorough analysis algorithm’s compared actual values.