Investigation of a ternary blend of diesel-pentanol-ethanol on engine performance, combustion and emission: A modelling and fuzzy optimization approach with artificial neural networks

Abstract The present study investigates the performance, combustion and emission characteristics of ternary blends of diesel, pentanol and ethanol through combined experimental and computational approaches. Pentanol is selected as a co-solvent for its unique chemical compatibility and ability to stabilize ethanol-diesel blends, thus preventing phase separation and enhancing blend performance. Experimental tests were conducted at a fixed engine speed (1500 rpm), compression ratio (17.5:1) and injection timing (23° bTDC) to evaluate the influence of blend composition under varying loads. The results show that, compared to neat diesel, the ternary blends yield increases in brake thermal efficiency (0.27-3.38%) and NOx emissions (5.56-32.01%), while exhaust gas temperature and smoke opacity decrease (0.73-6.1% and 4.59-22.57%, respectively). Artificial neural network (ANN) modeling, using a 4-16-4 architecture and Levenberg-Marquardt training, accurately predicts engine responses. The fuzzy logic system, applied to experimental and ANN data, identifies 85% diesel, 9% pentanol and 6% ethanol at 40% load as the optimal blend for performance-emission trade-off. This integrated approach demonstrates the effectiveness of artificial intelligence (AI)-driven optimization for advancing sustainable alternative fuels.