POD-RBF approximation in material characterization problems assisted by simulation models of the experiments

Sensitivity analysis and minimization procedures relevant to identification and optimization problems imply the repeated simulation of a system response with a few varied parameters among those defining the input data. This requirement fosters the development of methods capable of reducing the overall burden with no significant accuracy decay. Sophisticated but expensive computations, for instance carried out by non–linear finite element approaches, may be replaced by simplified analytical formulations based on the interpolation of some numerical results. Different interpolation schemes offer distinct approximation capabilities. Accuracy also depends on numerical disturbances, which can be filtered by data compression tools that retain the only essential features of the considered system response. This contribution summarizes some experiences with POD–RBF approximation in material characterization problems assisted by the numerical simulation of the experimental work