Models for the description of mechanical systems contain constant parameters, which, in general, are to be determined from experimental investigations. Therefore numerical optimization strategies are employed, which identify iteratively the best set of parameters for a given model. Thereby the target function, which is to be defined above, influences significantly the quality of the results. The hybrid strategy developed here combines the excellent global search characteristics of stochastic evolution strategies with the high rate of convergence of deterministic gradient and simplex methods.
Example: Determination of model parameters of a material model for bitumen by means of an evolution strategy
The evolution strategy employs genetic algorithms, e.g. selection, recombination or mutation, to optimize the parameters of the investigated material model.
Results from experimental and numerical investigations of tensile and creep experiments are presented, which have been taken into account passing the optimization procedure including evaluation of the target function. The depicted numerical results employ the optimized set of parameters.