The direct influence of mean flow on subgrid stresses in LES of turbulent flows

Abstract

Large eddy simulation (LES) presents an alternative to Reynolds-averaged turbulence modeling in the prediction of complex turbulent flow fields in engineering applications. In LES, the large scales are explicitly computed and models are provided for the effect of small scales on the computed large-scale dynamics. The expectation is that because the large eddy part of the turbulence is explicitly computed, the subgrid-scale (SGS) modeling can be simpler and perhaps more universal than classical turbulence models. The chapter investigates the anisotropy induced by mean flow gradients in the subgrid stresses and associated energy transfer. An analysis leads to the identification of a rapid SGS stress tensor that depends explicitly on the mean velocity gradient and a slow SGS stress tensor that depends only on the fluctuating velocity. Previous direct numerical simulation (DNS) of the mixing layer is then used to quantify the relative importance of the rapid term as well as explore implications for SGS models. A priori tests are performed using an existing direct numerical simulation database of a temporally evolving turbulent mixing layer. Quantitative and qualitative agreement between numerical results and the analysis are obtained.