Assessment of LES Subfilter Model Accuracy in a Supercritical CO2 Spatially Evolving Mixing Layer

The dynamics exhibited in a spatially evolving supercritical carbon dioxide mixing layer are explored from a numerical perspective through joint application of the direct numerical simulation (DNS) and large-eddy simulation (LES) methodologies. The DNS data is used as a reference from which an observational assessment of the accuracy of the LES calculations is made. The LES calculations are performed on three grids of varying resolution. A wall-resolved calculation is performed, as well as calculations on grids two successive levels coarser in the spatial and temporal domains. The mixed-dynamic Smagorinsky and gradient diffusion models are employed as the turbulence closures. Qualitative results presented include three-dimensional visualizations of the mixing fields based on the Q-criterion. The results highlight significant coupling between the thermodynamic and transport nonlinearities found at the selected flow conditions and the underlying turbulent field. Quantitative observations are made through one-dimensional line plots with data sampled at four different axial locations in the field. Instantaneous data is presented, along with ensemble averaged trends and standard deviations. The results broadly indicate that the LES calculations lose accuracy as axial distance downstream of the initial mixing location increases. Significant drops in accuracy are found between the wall-resolved calculation and the successively coarser resolution LES calculations, leading to insights into grid resolution requirements for accurate LES at these conditions.