Prediction of performance and exhaust emissions of diesel engine fuelled with adulterated diesel: An artificial neural network assisted fuzzy-based topology optimization

This study evaluates the effects of diesel fuel adulteration on the performance and exhaust emission characteristics of an existing diesel engine. Kerosene is added to diesel fuel in volumetric proportions of 5, 10, 15, and 20%. Adulterated fuel significantly reduced the oxides of nitrogen emissions of the engine. In view of the engine experimentations, artificial intelligence-based artificial neural network model has been developed to accurately predict the input-output relationships of the diesel engine under adulterated fuel. The investigation also attempts to explore the applicability of fuzzy logic in the selection of the network topology of artificial neural network model under adulterated fuel. A (2-7-5) topology is found to be optimal for predicting input parameters, namely load, diesel-kerosene blend and output parameters, namely brake thermal efficiency, brake-specific energy consumption, oxides of nitrogen, total hydrocarbon, carbon monoxide of the network. The developed artificial neural network model is enabled for predicting engine output responses with high accuracy. The regression coefficient (R) of 0.99887, mean square error of 1.5e-04 and mean absolute percentage error of 2.39% have been obtained from the plausible artificial neural network model.