Model Based Determination of Dynamic Engine Control Function Parameters

The use of model based approaches in areas such as simulation, control design, optimization, etc. is crucial for the development of highly sophisticated systems. This is especially true for typically very tight time-to-market frames. Physical modeling of IC engine emissions based on first principles is extremely complex and still requires by far too much calculation time. However, special fast neural networks represent a promising alternative for an accurate modeling of the emission behavior, even for dynamic conditions.

This paper first describes the process of developing dynamic neural emission models. The required data is collected by a specially designed dynamic measurement strategy. The models themselves are then used for the optimization of the dynamic engine behavior concerning consumption, emissions and drivability.