Traditional uncertainty quantification in simulation-based analyses typically focuses upon the quantification of parametric uncertainties: the uncertainties inherent to the input parameters of the simulation. This type of uncertainty often represents the inherent variability of natural systems due to incomplete information or implicit variations in the system itself. However, when simulation-based models are used to evaluate the performance of a system, an additional level of uncertainty is introduced (one often not rigorously considered in probabilistic design): modeling-induced uncertainty. This work adapts uncertainty quantification for the multiple forms of uncertainties into a single framework to quantify the uncertainty in an analysis from not only the inherent variability of the parameters but also the uncertainties induced by the modeling process itself. By using Bayesian statistics, a stochastic representation of model results can be constructed in the presence of given data. These data can then be used to drive rigorous reduction and selection of models among a model set that most accurately represent the given physical scenario. The utility and application of such an approach will be demonstrated on a transonic aeroelastic flutter analysis for the Advisory Group for Aerospace Research and Development 445.6 wing.
Quantification of Modeling-Induced Uncertainties in Simulation-Based Analyses
AIAA Journal ; 52 , 1 ; 195-202
01.01.2014
Aufsatz (Konferenz) , Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
Quantification of Modeling-Induced Uncertainties in Simulation-Based Design
British Library Conference Proceedings | 2012
|Evidence-Based Quantification of Model-Form Uncertainties in Simulation-Based Analyses
British Library Conference Proceedings | 2013
|