Abstract
This paper explores the feasibility of using engine operating parameters to predict exhaust emissions from a direct injection diesel engine through the use of neural networks. In a first stage, artificial Neural Networks are presented, focusing their basic working principles and the cases in which this tool is of some use. In a second step, these theoretical concepts are applied to a more concrete problem: the prediction of the oxides of nitrogen (NOx) generated in a diesel engine. The analysis of the neural network information can lead to some conclusions about the importance of some of the engine operating parameters in the final NOx production. In addition, once the nets are fully trained, predictive emission maps with respect to the engine parameters can be generated.