Sensitivity computation for uncertain dynamical systems using high-dimensional model representation and hierarchical grids

Abstract

Global sensitivity analysis is an important tool for uncertainty analysis of systems with uncertain model parameters. A general framework for the determination of sensitivity measures for fuzzy uncertainty analysis is presented. The derivation is founded on the high-dimensional model representation, which provides a common basis with Sobol indices, illustrating the similarities and differences of fuzzy and stochastic uncertainty analysis. For the numerical calculation, a sparse-grid approach is suggested, providing an efficient realization due to the direct relationship between hierarchical grids and the sensitivity measures.