Workshop on Uncertainty Quantification in Computational Science
June 11-12, 2007
Achieving predictive simulations of physical systems requires a concerted effort in verification, validation, and uncertainty quantification (UQ), including rigorous assessment of model/code validity through comparisons against experimental measurements, with well-characterized uncertainty/error bars for both experimental and computational results. This workshop will present a number of UQ techniques, focusing on generalized polynomial chaos expansions (GPCE) to represent random variables and processes, as well as various techniques for uncertainty propagation in systems governed by ordinary or partial differential equations, with applications in chemistry, thermofluids, materials, etc.
Some specific topics to be addressed include
- Galerkin modeling in stochastic spaces, including computational solution aspects, error estimation, and post-processing.
- Non-intrusive (sampling-based) and intrusive (direct) UQ methods.
- Bayesian methods for estimation of uncertain parameters from data.
- Current research topics, including interfacing multiscale and stochastic modeling, GPCE and Bayesian based stochastic optimization for stochastic partial differential equations, and UQ for oscillatory dynamical systems.