912 Advances in Uncertainty Quantification for Multi-physics Applications

Tim Wildey, Sandia National Labs
Tan Bui, The University of Texas at Austin
Josh Shadid, Sandia National Labs
Many problems in science and engineering are best described by multi-physics models that interact on a wide range of length and time scales and are subject to various sources of uncertainty due to unknown material properties, boundary conditions, etc. In order to integrate data in a consistent manner and to make credible predictions for quantities of interest with quantified error and uncertainty, in recent years, the computational science and engineering community has sought to develop mathematical and numerical tools to enable the incorporation of information from all relevant spatial and temporal scales in a simulation. In many cases, the goal is to provide accurate estimates of certain probabilistic predictions (e.g., mean, variance, probability of failure, etc.) for a handful of quantities of interest. This task is especially difficult for large-scale multi-physics applications where the number of uncertain parameters may be quite large, the number of high-fidelity model evaluations may be limited, and the available data may be corrupted by significant noise. Some recent advances have sought to adaptively control both the deterministic and stochastic sources of error using a posteriori error estimates, while other approaches seek to utilize lower-fidelity models, such as surrogate or reduced-order models, within a multi-level or multi-fidelity framework to reduce the computational complexity.

The goal of this mini-symposium is to provide an opportunity for researchers to present recent work and exchange ideas on novel methods for forward and inverse propagation of uncertainty and applications of these techniques to challenging multi-physics applications.

We anticipate contributions on the following topics:
- Efficient methods for forward and inverse propagation of uncertainty for multi-physics applications
- Multi-level and multi-fidelity methods
- Novel approaches for solving stochastic inverse problems
- A posteriori and a priori error estimates that separate deterministic and stochastic sources of error
- Combined spatial-temporal-stochastic adaptive techniques