Analysis of the near-wall behaviour of some self-adaptive subgrid-scale models in finite-differenced simulations of channel flow

Self-adaptive subgrid-scale models are proposed and assessed. They are based on the use of the Germano-Lilly dynamic procedure and the use of a selection function. These models, which do not incorporate any information related to the location of the solid walls, are well suited for the simulation of turbulent flows in complex geometries. Their reliability, when used together with a second-order non-dissipative. numerical method, is assessed on the plane channel configuration for two values of the Reynolds number (Re-tau = 180 and 395) for two grid resolutions. The selection function approach for deriving self-adaptive subgrid models is found to yield results very similar to those obtained using a dynamic model, without requiring any numerical stabilization procedure. The use of the selection function is shown to be the only one which is able to capture the backscatter process in the buffer layer, while producing a strictly positive subgrid viscosity. This is demonstrated to be linked to the capability of the selection function to permit a decorrelation between the mean strain and the fluctuations of the subgrid stresses. That point is illustrated thanks to the introduction of a new decomposition of the fluctuating strain subgrid dissipation. Copyright (C) 2002 John Wiley Sons, Ltd.