Application of Artificial Neural Networks in Exhaust Gas Recirculation Systems

Nitrogen oxides are currently one of the most serious problem of atmospheric pollution and the ‘low emission’. One of the most important source of emissions of these pollutants is road transport. Therefore, it is necessary to develop the most effective methods for reducing the formation of these substances in internal combustion engines.

In the article, the possibilities of using artificial neural networks in exhaust gas recirculation systems are presented. One of the advantages of using artificial neural networks is the possibility of using to identify control systems in which the mechanism of processing input signals into their corresponding output signals is unknown. This article presents the use of neural networks to control exhaust gas recirculation. The control has been based on input signals from the mass air flow sensor, rotational speed and load value, which indirectly reflected the fuel dose. The output parameter was the adjustment of the EGR valve opening in such a way that a constant air flow in the intake system was maintained for the given input parameters. For the input parameters defined in this way, measurements were taken at the test bench. The data acquisition from the measurements carried out was used to build the model of artificial neural network. To verify its correctness, comparative tests of harmful and toxic exhaust emissions were carried out for valve control by the EDC unit and for a valve controlled by an electronic system using artificial neural networks.