Reducing the Dependence of Optimal Finite-Volume LES Models on DNS Data

The Optimal finite-volume large-eddy simulation (LES) approach has successfully been applied to the forced isotropic turbulence case. This modeling approach is based on the stochastic estimation of finite-volume fluxes, with the numerical Navier-Stokes and subgrid terms being treated simultaneously. Previously, the stochastic estimation procedure required direct numerical simulation (DNS) data to supply the necessary correlations. For this modeling approach to be generally applicable, DNS data cannot be used. Two alternate sources for the correlation information have been explored. The first makes use of inertial range theory and the approximations underlying EDQNM to yield infinite Reynolds number models. A second approach, which may be called a phdynamic model, relies on velocity data from an ongoing LES simulation to compute most of the correlations. Results for both approaches will be described, with the strengths and shortcomings of the analytical and dynamic methods pointed out.