Sustainable biodiesel from Mesua ferrea seed oil using explainable machine learning for engine performance evaluation...

Area of Science:

• Renewable Energy
• Combustion Engines
• Environmental Science

Background:

• Growing demand for sustainable energy solutions and net-zero emission targets drives research into alternatives for fossil fuels.
• Biodiesel from non-edible sources presents a viable option for reducing greenhouse gas emissions and leveraging existing engine infrastructure.
• Mesua ferrea seed oil is explored as a non-edible feedstock for sustainable biodiesel production.

Purpose of the Study:

• To experimentally investigate the engine performance and emission characteristics of biodiesel derived from Mesua ferrea seed oil.
• To evaluate the impact of varying fuel injection pressures and engine load conditions on biodiesel combustion.
• To apply and validate explainable machine learning models for predicting engine performance and emissions.

Main Methods:

• Physicochemical characterization of Mesua ferrea seed oil biodiesel.
• Experimental testing in a single-cylinder diesel engine under varied load conditions and fuel injection pressures (200-240 bar).
• Application of machine learning algorithms (XGBoost, Linear Regression, Decision Tree) with Explainable AI (XAI) validation using SHAP and PDP.

Main Results:

• Increased injection pressure positively impacted brake thermal efficiency and brake specific fuel consumption.
• Higher injection pressures led to significant reductions in Carbon Monoxide (CO) and Hydrocarbon (HC) emissions.
• Nitrogen Oxide (NOx) emissions exhibited an increasing trend with rising in-cylinder temperatures.
• XGBoost model demonstrated superior predictive accuracy (R²_test: 0.932-0.969) for engine performance and emissions.
• Explainable AI analysis highlighted the dominant influence of engine load on BTE, BSFC, and NOx, and fuel injection pressure on CO and UHC.

Conclusions:

• Biodiesel from Mesua ferrea seed oil is a promising renewable fuel for compression ignition engines.
• Optimizing fuel injection pressure is crucial for enhancing engine efficiency and mitigating specific exhaust emissions.
• Explainable machine learning provides robust tools for understanding and predicting the complex relationships between operating parameters, engine performance, and emissions.