Multifidelity uncertainty quantification with models based on dissimilar parameters

Multifidelity uncertainty quantification (MF UQ) sampling approaches have been shown to significantly reduce the variance of statistical estimators while preserving bias highest-fidelity model, provided that low-fidelity models are well correlated. However, maintaining a high level correlation can be challenging, especially when depend on different input uncertain parameters, which drastically reduces correlation. Existing MF UQ do not adequately address this issue. In work, we propose new strategy exploits shared space improve among with dissimilar parameterization. We achieve by transforming original coordinates onto an auxiliary manifold using adaptive basis (AB) method (Tipireddy and Ghanem, 2014). The AB has two main benefits: (1) it provides effective tool identify low-dimensional each model represented, (2) enables easy transformation polynomial chaos representations from high- spaces. This latter feature is used without requiring additional evaluations. present algorithmic flavors estimator cover analysis scenarios, including those legacy non-legacy high-fidelity (HF) data. provide numerical results for analytical examples, direct field acoustic test, finite element nuclear fuel assembly. For all compare proposed against both single-fidelity based