Meta-modeling or surrogate-modeling is a process to win the mathematical relationship between design parameters and product characteristics. For each point in the parameter space, there is an corresponding point of the design space. Many model calculations should be performed to show the relationship between outputs and inputs systematically (Full Factorial Design). For a high computing effort of the product model, it is practically infeasible. Adaptive response surface methodology can be used to solve this problem. First of all, predefined support points in the parameter space are calculated by the original model. Then, a response surface will be build using interpolation between these support points. Based on this response surface, new support points in the parameter space will be suggested and calculated by the original model. Closely, a new response surface will be build by all existing support points. This adaptive process will be continued, until either a defined accuracy of the response surface or a defined number of model calculations have been reached.
The mathematical relationship between design parameters and product characteristics presents a new dimension of the simulation results and it is so-called meta-model. Based on this meta-model, a virtual optimization or test of the virtual design can be performed very fast to evaluate and to improve the design under real conditions. The meta-model can be exported into C-, Visual Basic, Modelica-, or Matlab-code for further using as surrogate model in system simulation (Matlab/Simulink, SimulationX, Dymola etc.)