CSE Community Seminar
September 27, 2024, 12-1PM
Conference Room 45-432 in Building 45
Physics-guided uncertainty quantification of RANS turbulence model for mixed convection flows
Yu-Jou Wang
Postdoctoral Associate
Department of Nuclear Science and Engineering, MIT
Abstract
In this study, we propose a framework for quantifying the model uncertainty in Unsteady Reynolds Averaged Navier-Stokes (URANS) simulations under transient conditions. Within this framework, the model uncertainty is treated as a Gaussian random field and injected through the turbulence viscosity; a dynamic turbulence-tracking approach is adopted, where the spatiotemporal variability of the covariance is guided by local turbulence characteristics. By constraining model uncertainty with the history of flow evolution, this approach offers a robust and efficient strategy for estimating the time-dependent model error, eliminating the need to evaluate all possible solutions at each time step exhaustively.
The capability of the proposed framework is demonstrated in the Supercavna transient experiment, in which the flow scenario is an approximate representation of the complex flow in the upper plenum of sodium fast reactors (SFR). Three URANS models, realizable k-ε, nonlinear eddy viscosity k-ε with a cubic stress-strain relation, and Wilcox k-ω model were tested. Results show that the proposed framework can dynamically identify the regions where large sources of error exist and provide a good estimation of model uncertainties.
September 27, 2024, CSE Community Seminar
Yu-Jou Wang
Postdoctoral Associate
Department of Nuclear Science and Engineering, MIT