Neural network for modeling and optimization of internal combustion engines

In this work, the possibility of modeling the engine emissions has been analyzed in terms of NOx, CO, unburned hydrocarbons and brake specific fuel consumption as a function of injection parameters and EGR levels. The results showed that the neural network technique was capable of analyzing the influence of any single input parameter on any output variable and allowed the formulation of an efficient simple mathematical model of the engine that was able to predict the output of interest.