Model-based map calibration on combustion engines

Engine developers face legal low emission rates and the demand for less fuel consumption by implementing complex functionalities to optimize the combustion process. Counteracting the extremely rising costs in calibrating the electronic control unit (ECU), this paper presents a new multi-step approach. In the first phase, design-of-experiment (DOE) methods are used to transform the calibration task into few experiments. In phase 2 this list for the engine test bed yields to the database for the modeling in phase 3. For system identification multivariate regression (MulReg) and artificial neural networks (ANN) are used. Applied to this models, optimization at the operating points leads to list of theoretical optima-candidates, which have to be verified in phase 4. The final map calculation in phase 5 produces the new ECU maps. A practical application represents this method.