Subgrid-Scale Models based upon the Second-Order Structure-Function of Velocity

Abstract

Large-eddy simulations of wakes and boundary layers are performed with the aid of the structure-function model [[1]] and two of its sequels, namely the filtered and selective structure- function models. It is first checked, in the case of an incompressible wake, that the original structure-function model indeed enables one to obtain higher Reynolds-number dynamics than in direct simulations at the same resolution. This is also verified with the Smagorinsky model, but with a slightly lower level of three-dimensionality. The structure-function model is then used to simulate an incompressible wake developing spatially.

Both the Smagorinsky and the structure-function models are too dissipative to simulate transition of weakly compressible boundary layers, at least at a reasonable cost (say, less than 50 hours of Cray 90). This has been overcome by simply filtering out the large scales before computing the eddy-viscosity (filtered structure-function model). Another option is proposed: apply the structure-function model only when and where the spatial variation of the orientation of the vorticity vector exceeds a certain threshold (selective structure-function model). This also works, with results very similar to those obtained with the filtered structure-function model.