A neural network approach for the correlation of exhaust emissions from a diesel engine with diesel fuel properties

This paper presents expressions correlating the exhaust emissions from a single-cylinder diesel engine with some of the most important properties of the fuels used, using a neural network approach. The exhaust emissions measured were carbon monoxide, hydrocarbons, nitrogen oxides, and particulate matter. The experiments were performed using a matrix of 59 fuels. The cetane number of the fuels covered the range 42-58, the density varied between 0.840 and 0.860 g/mL, and the sulfur content from 0.05 to 0.20 wt %. The predictions were based on specific points of the distillation curve, the cetane number, density, and kinematic viscosity of the fuels. In the case of particulate matter emissions, sulfur content was also employed. The predictions obtained were very good for all the emissions considered. The aromatic content was not used as a predictor variable, because it was found to have a strong inter-correlation with the cetane number, density, and two specific points of the distillation curve, the 50% and the 90% recovery point.