Efficient Utilization of Surrogate Models for Uncertainty Quantification

Numerical simulations for the analysis and design of structures or systems are often based on deterministic characteristics, whereas reality is determined by data information which characterized various types uncertainty (variability, imprecision, inaccuracy, incompleteness). Besides traditional probabilistic approaches, possibilistic models most recently in focal point research. Combining characteristics aleatoric epistemic uncertainties, polymorphic yields approaches related to application field imprecise probability models. Uncertainty schemes, pointwise evaluation a fundamental solution (structural analysis), usually lead high computational costs, due repetitive evaluations. Especially complex models, each characteristic demands separate quantification. Hence, surrogate indispensable Considering complex, dimensional structural even single model evaluations defined significant cost. Therefore, transition from space-filling sampling schemes towards adaptive allows an overall decrease cost while maintaining necessary accuracy prediction. A variance strategy proposed, constituted combination exploration exploitation given input space. The predictions' variances multiple hereby utilized detect regions interest define points accordingly. contribution briefly introduces fundamentals fuzzy-set theory interval as well model. generalized framework presented with special respect corresponding effort. Subsequently, concept scheme explained further multi story example uncertain material properties.