Application of Neural Networks for Prediction and Optimization of Exhaust Emissions in a H.D. Diesel Engine

A study of the feasibility of using engine operating parameters to predict and minimise exhaust emissions from a direct injection H.D. Diesel engine through the use of Neural Networks (NN) was conducted.

The objective is to create a mathematical tool that, learning from a large number of experimental data obtained under different operating conditions, is able to parametrize oxides of nitrogen (NOx) and particulate matter (PM) exhaust emissions as a function of engine operating parameters.

Once satisfactory NN predictive results were obtained, the tool was also used to simultaneously optimise several operating parameters for low exhaust emissions. The optimisation was based on a minimising process related to EURO IV standards regulations.