Innovative Methodologies for Robust Design Optimization with Large Number of Uncertainties Using ModeFRONTIER

A Bayesian approach for quantile optimization problems with high-dimensional uncertainty sources

“…The collaborative EU research project UMRIDA [16] (Uncertainty Management for Robust Design in Aeronautics, 2013-2016), tackled robust design under a large number of independent uncertainties. In this project, problems with more than 10 to 20 independent uncertain parameters were considered high dimensional and treated by dimensionality reduction techniques such as Karhunen-Loeve expansions [17,18], and sparse Polynomial Chaos [19,20] expansions. The construction of surrogates in high dimensions can also benefit from the gradients, estimated, for instance, by adjoint methods [21].…”