Sparse-Grid Surrogate Models for Uncertainty Analysis

The computation of fuzzy arithmetical solutions of problems involving uncertain, fuzzy-valued model parameters can be formulated as a nested sequence of optimization problems. In order to reduce the amount of model evaluations, which is often a limiting factor for the applicability of uncertainty analysis if the original model is evaluated, a surrogate-model approach based on sparse-grid interpolation is investigated. The optimization process is then performed on the basis of the surrogate representation, leading to a significant improvement with respect to the computation time. (© 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim)