MULTILEVEL QUASI-MONTE CARLO FOR INTERVAL ANALYSIS

This paper presents a multilevel quasi-Monte Carlo method for interval analysis, as computationally efficient high-dimensional linear models. Interval analysis typically requires global optimization procedure to calculate the bounds on output side of computational model. The main issue such is that it numerous full-scale model evaluations. Even when simplified approaches vertex are applied, required number evaluations scales combinatorially with input intervals. increase in especially problematic highly detailed numerical models containing thousands or even millions degrees freedom. In context probabilistic forward uncertainty propagation, multifidelity techniques show great potential reduce cost. However, their translation an not straightforward due fundamental differences between and methods. this work, we introduce framework. First, intervals transformed Cauchy random variables. Then, based these variables, sampling designed. Finally, corresponding responses post-processed estimate quantities high accuracy. Two examples technique very medium comparison traditional propagation results well within predefined tolerance.