I am now using stagger grid to compute channel flow in which streamwise and spanwise velocity component (u,w) points are at the same levels,normal velocity component(v) points are at the inter-levels.The first u grids locate very near but not at the wall(y+<5).How can I set the condition for this case? There are many cases using log-law,but I want to resolve the flow very near wall. Is there any way to give SGS Reynolds stress at the wall?

Thanks!

Hi there,

To impose no slip on the wall, you can either use ghost cell and then insure that the interpolation of the velocity on the surface is zero, or you can use a "zero-volume" cell for the boundary, and then automatically you have U=v=w=0.0 at the wall. Hope this helps. Sincerely,

frederic Felten

PS: if you need more info, just check the book by Ferziger and Peric, 1996.

Do you calculate SGS properties directly?

In LES, only GS is calculated from filtered N-S equations. SGS properties are modeled (or calculated) from GS properties. As far as I know, you calculate GS quantities such as filtered velocities, pressure and in this case you need GS properties as boundary conditions on wall (If your governing equations have only GS components.)

As Mr. Frederic mentioned, you can use no-slip boundary condition on the wall, i.e. v=0, u and w is calculated from wall model(In your staggered grid). Please draw your control volume at wall boundary and place velocity components and pressure. Maybe you can find your solution easily.

Jongdae Kim

Thanks alot!

In my case,the grids near the wall are:

U,W,P level ______________________ V level,wall(y=0) ______________________

-1 ______________________

According to many reference,wall model requires that the first U grid lie in log-law region.But in my case It is where y+<5.Maybe the ghost cell is workable.Actually,I have tested this method,and the velocity profile is good as well as r.m.s of velocity fluctuations,but the skewness and flatness results are very bad.SGS is modelled,but near the wall, many guys use wall model instead of boundary condition of velocity.

What I wonder is if ghost cell is used,in which only viscous stress is given without SGS Reynolds stress component,whether the former is enough to model all the wall effects. Is the latter needed to put another way like wall model to take into account?