Evolutionary Algorithms for the Automatic Calibration of Simulation Models for the Virtual Engine Application

The automotive industry is, based on the sales, the most important industry section in Germany and other countries. Because of many various boundary conditions, like CO2 and fuel reduction, the calibration of the engine control unit is growing up for the development of the whole system automotive. Traditional calibration methods fail because of the increasing complexity of the optimization tasks and the huge measurement effort. A one dimensional engine process simulation in combination with a physical combustion model can significantly contribute to essential parts of engine calibration. For an accurate reproduction of the reality the simulation model must be calibrated. Factors are used as calibration parameters, which take influence on various combustion values. With the help of an automatic optimization routine the time effort can be reduced significantly for the adjustment with constant or increasing accuracy. Therefore, the main target of this contribution is the improvement of the simulation and optimization techniques. For this aspect an evolutionary algorithm, which is a combination of a genetic algorithm and an evolutionary strategy, is used for the optimzation task. Evolutionary optimization methods are stochastic searching methods, which are leaned on the naturally biological evolution. They work simultaneously with a number of potential solutions and they are variable, robust and powerful. They give the opportunity to solve complex and multi-criteria problems in a reasonable time. Further advantages of this approach are the better results and the time saving. The presented evolutionary algorithm for the automatic calibration of simulation models for the virtual engine application has been evaluated by simulations.