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

Area of Science:

• Sustainable energy sources
• Combustion engineering
• Machine learning applications in fuel science

Background:

• Biodiesel is a sustainable alternative fuel with lower emissions and engine compatibility.
• Rising crude oil prices necessitate exploration of viable biodiesel feedstocks.
• Hydrogen-diesel blends offer potential for enhanced engine performance and reduced environmental impact.

Purpose of the Study:

• To evaluate multiple biodiesel types (Mustard, Binola, Caster, Rapeseed, Canola, Peanut, Jatropha, Thumba) for use in hydrogen-diesel blends.
• To determine the optimal biodiesel for improved engine performance (BTE, BSFC) and reduced emissions.
• To integrate experimental data with machine learning for robust biodiesel blend analysis.

Main Methods:

• Experimental analysis of engine performance characteristics (BTE, BSFC) and emissions.
• Application of k-means clustering and prediction models for biodiesel optimization.
• Validation of biodiesel performance against ASTM standards.

Main Results:

• Canola biodiesel demonstrated superior performance with 31.2% BTE and 0.26 kg/kWh BSFC.
• Canola biodiesel achieved significantly reduced emissions: 0.03% CO, 24 ppm HC, 32 mg/m³ PM, 4.5 ppm SO₂, and 770 ppm NOx.
• Machine learning models identified Canola as the optimal biodiesel with high accuracy (RMSE=1.0, MAPE=2.95%, R²=0.95).
• Canola and Rapeseed oils formed the best-performing group.

Conclusions:

• Canola biodiesel is the most promising sustainable fuel option among those tested.
• Favorable fatty acid composition in Canola oil contributes to its superior combustion and performance.
• The hybrid machine learning approach provides a robust method for selecting optimal biodiesel blends.