Improved control of electronic diesel injection by means of neural networks

The search of more power and softness, as well as bigger efficiency in the combustion, smaller fuel consumption and emission of polluting gases in the current engines, demands more and more complex and critical control systems. In this work an improvement of the electronic injection, based at the moment on cartographic methods, by means of the application of neural networks is presented. The neural networks allow to adjust the values obtained to each exact point of operation, forgetting the intermediate steps of the previous methods or another approach that are not adjusted to the ideal function. In addition the proposal reduces considerably the number of data that are necessary to store for very exact controls based on cartographies pre-calculated. On the other hand this approach opens the doors to intelligent systems in the own injection control, that by means of self-learning, will obtain improvements in the methods to obtain the data of the engine operation and their direct and instantaneous interaction on the engine. In this manner, it is possible to achieve that the injection be accurate during all the engine lifetime.