Determining Model Form Uncertainty of Reduced Order Models

摘要

The quantification of model form uncertainty is very important for engineers to understand when using a reduced order model. This quantification requires multiple numerical simulations which can be computationally expensive. Different sampling techniques, including Monte Carlo and Latin Hypercube, are explored while using the maximum entropy method to quantify the uncertainty. The maximum entropy method implements random matrices that maintain essential properties. This is explored on a planar frame using different types of substructure representations, such as Craig-Bampton. Along with the model form uncertainty of the substructure representation, the effect of component mode synthesis for each type of substructure representation on the model form uncertainty is studied.