LES: boundary layer computation

Francois

Hi all,

I'm using LES in fluent with the Dynamic Samgorinsky model to compute the boundary layer flow of an incompressible fluid past a deep cavity of length L and deep D.

___<--L--> ____

^

|

D

|

^__________

The computational domain is as follow:

_____________symmetry________________

inlet.......................................pressu re outlet

__wall________cavity_____wall_________

The mesh is fine enough: y+ <1 on all the wall on the bottom of the computational domain.

On the upper part of the domain, symmetry boundary condition is used (at 3 D from the wall). At the outlet pressure outlet is used (at 4 D downstream the cavity). At the inlet (at 3 D upstream the cavity) the fully developped velocity profile comes from experiment.

The transverse (Oz) direction is homogeneous, so peridodic boundary are used in this direction with:

* a pressure gradient fixed to zero * flow direction fixed to x=0,y=0,z=1 * the periodic boundary condition are translational type

Convergence works fine but the velocity computed in the boundary layer upstream of the cavity don't match the fully devlopped profile of a turbulent boundary layer which is already injected at the inlet. And the skin friction coefficient decreases upstream from the cavity but in the experiment it's constant.

What's wrong with this configuration ? I have tried with 3D RANS modeling and it seems to work (for K-e for exemple) ...

Is the domain too short upstream from the cavity for an LES simulation ?

Is the Dynamic Smagorinsky model not appropriate ?

Are the periodic boundary conditions note appropriate for this configuration ?

Thank's for your help ! Francois