Roughness length, interpolation, unphysical dissipation rate on the ground
 

Hello to everybody,

I am trying to model a very simple 3-D problem (an empty wind tunnel) with hexahedral cells and I have to apply an experimental wall roughness on the ground equal to 1.53m (full scale).
I have used inlet and outlet boundary conditions, symmetry for the lateral planes, inlet at the top boundary (with constant u, k, epsilon), fully rough wall on the ground.
Standard k-epsilon High Reynolds turbulence model used with constant profiles for u, k and epsilon derived from the experiment.

In star-cd when standard wall functions used for fully rough wall I have two options according to my knowledge: the equivalent sand grain roughness and the roughness length, and it depends on the format of the equation u=u*/κ ln(z/z0) + A. If A=0 then z0 is the roughness length.
Also the centre of the first cell near the wall has been selected about 4 times the z0 an extremely high value but it is consistent with the manual.

When the steady state simulation finished, I noticed that near the outlet of the domain and inside the first cell near the ground an unphysical behaviour for epsilon using sensors at three positions. One on the ground z=0 where epsilon was zero (?) while it should be infinite, at the centre of the cell (z=dz/2) epsilon had a low value and at the edge of this cell (z=dz) the dissipation value was exactly the same with what I gave at the inlet i.e. the correct one.

Do you think that this is normal to happen? Is it a boundary condition problem or an interpolation problem or a user problem?
Are there any alternative ways to insert the roughess length e.g. using other near-wall treatment?

Thank you in advanced

Kind regards

Giorgos