A surrogate-model-based identification of fractional viscoelastic constitutive parameters

Abstract

In order to reduce the computational expense, a Kriging surrogate model is developed as an approximation of a numerical model based on FEM (finite element method) and FDM (finite difference method) to solve direct fractional viscoelastic problems and then is combined with a gridding-partition-based continuous ant colony algorithm to identify constitutive parameters of fractional viscoelastic materials. Three kinds of modeling strategies are presented to generate the Kriging surrogate model, that is, global modeling, piecewise modeling, and reduced modeling. Two numerical examples are given to illustrate the proposed approach in terms of computing accuracy and expense. The utilization of Kriging surrogate model not only can provide a sufficient computing accuracy, but also can significantly reduce the computational cost in solving inverse fractional viscoelastic problems. In addition, regional inhomogeneity and impact of noisy data are taken into account.