Use of kriging to surrogate finite element models of bonded double cantilever beams

An algorithm based on kriging statistical interpolation for computing the surrogate response of a Finite Element model is presented. The interpolation model is calibrated via computation of Finite Element responses at a set of random occurrences of a material parameter. Such random generation concentrates at locations where the numerical model requires a higher amount of data to get the desired accuracy. As a model problem a standard fracture propagation test is analyzed. The proposed procedure is shown to be robust and accurate since responses obtained via a direct computation and use of the surrogate model turn out to be indistinguishable.