Abstract
The increasing global cost of fossil fuels and growing environmental concerns have accelerated the search for sustainable energy alternatives, positioning bioethanol as a promising renewable fuel for spark-ignition (SI) engines. This study uniquely integrates ethanol-petrol blends (E0, E10, E20, and E30) with Artificial Neural Networks (ANNs) to address critical gaps in predictive modeling and fuel optimization. Experimental tests were conducted on a single-cylinder, four-stroke SI engine under constant load conditions, capturing data on engine speed, mass flow rate, combustion efficiency, peak cylinder pressure, brake-specific fuel consumption (BSFC), and exhaust gas temperature. A feed-forward backpropagation ANN model was developed using 75% of the collected data for training and 25% for validation, achieving high predictive accuracy with R(2) values exceeding 0.98 for most parameters. Results showed that E30 improved combustion efficiency by 12.5% compared to E0 at 1500 RPM and reduced BSFC by 22% in the 2000-2500 RPM range, while maximum cylinder pressure increased with RPM but remained slightly lower for higher ethanol blends due to ethanol's cooling effect. By effectively predicting performance metrics across a broad RPM range (1500-3500), the ANN model reduces reliance on extensive experimental testing and offers a scalable approach for optimizing fuel-blending strategies, thereby supporting the transition to cleaner, more efficient energy systems.