Exploring multi-feedstock biodiesel-hydrogen synergies for enhanced diesel engine performance using hybrid AI...

In recent years of crude oil shortage and price rise, biodiesel has emerged as a new-age sustainable fuel, owing to its positives mainly providing lower amount of gas emissions and compatibility with prevailing petro-diesel engines. The contemporary study explores multiple feasible biodiesels (Mustard, Binola, Caster, Rapeseed, Canola, Peanut, Jatropha, and Thumba) based on composition and testing feasibility, so as to be inter-mixed with hydrogen gas, for improved engine performance and minimal effluents. Primary engine characteristics brake thermal efficiency (BTE), brake specific fuel consumption (BSFC), and multiple effluents of diesel engine were experimentally explored and compared for specific fuel sets. A hybrid methodology comprising of machine learning (k-means clustering) and prediction models is employed to find the optimal biodiesel among the variants, validated against ASTM standards. BTE and BSFC show a strong negative correlation, CO and BSFC a positive correlation, CO₂ and BTE a positive one. With minimal prediction uncertainty, Canola-based oil emerged as the superior biodiesel among others with a RMSE = 1.0, MAPE = 2.95%, R² = 0.95, owing to favorable fatty acid composition. Canola and Rapeseed oils are placed in the best performing group, validating the improved performance efficiency characteristics and reduced pollutant output. Canola biodiesel achieved the best overall performance with 31.2% BTE, 0.26 kg/kWh BSFC, and low emissions of 0.03% CO, 24 ppm HC, 32 mg/m³ PM, 4.5 ppm SO₂, and 770 ppm NOx, due to improved atomization, oxygenated combustion. This study's novelty lies in integrating experimental data with hybrid machine learning models for robust of biodiesel blends.