Reacting Flows UQ
Uncertainty Quantification in Reacting Flows
Both the structure and parameters of reacting flow models rely on a significant degree of empiricism, or on approximate theoretical analysis, and therefore involve a degree of uncertainty. This is true for constitutive laws, subgrid models, initial/boundary conditions, as well as thermochemical models. In particular, chemical kinetic models for complex fuels typically involve thousands of uncertain Arrhenius rate-expression parameters.
It is important, both for model validation and design optimization purposes, that these uncertainties be adequately quantified and propagated to model outputs of interest. Sandia’s computational researchers have developed probabilistic algorithms and software tools to estimate uncertainties in model parameters based on measurement data, and to propagate these uncertainties forward through complex chemically reacting flow models. The former relies on Bayesian inference to construct joint posterior probability density functions (PDFs) on model parameters, while the latter relies on polynomial chaos expansions to represent random variables along with both intrusive/nonintrusive means to propagate these expansions forward. Moreover, the use of Bayesian inference strategies has been demonstrated in constructing Arrhenius rate expression parameter PDFs in the absence of data, but using other available information based on prior measurements.
The coupling of advanced uncertainty quantification methods and high-end computational resources enable the extraction of additional information from reacting flow computations that is useful in both scientific and engineering contexts.