Evaluation of Subfilter Model Performance for Large-Eddy Simulations of Supercritical Fluids

Abstract

The accuracy of three large-eddy simulations (LES) is assessed using a reference dataset obtained via direct numerical simulation (DNS). All of the LES simulations employ the dynamic Smagorinsky model to close the momentum equation and a dynamic gradient model to close the total energy equation. The LES data are obtained on three grids with resolutions spanning from the wall-resolved LES limit to two successive levels coarser in the spatial and temporal domains. The configuration employed for the study is a three-dimensional spatially evolving turbulent shear layer. The working fluid is pure carbon dioxide. The system is maintained at a supercritical state near the critical point such that the field is dominated by strongly nonlinear thermophysics. This allows the analysis to occur under conditions where the subfilter closures are significantly strained by the thermodynamics. Results explore characteristics of the turbulence from both a modeling and fundamental perspective. First, mixing layer growth rates are quantified. Discrepancies are found between the reference DNS data and the LES data. Energy spectra, turbulent transport coefficients, and Reynolds stress anisotropy results are presented to explore the origins of this mismatch.