[37269]

I have a long, straight (water) flume with a width of 0.10 m. The water depth is around 0.12 m for the case I'm testing.

According to hydraulic theory (and our lab verification), the vertical velocity profile should have a "dip", in the profile, instead of being a standard logarithmic one. (due to secondary currents). This is illustrated in the attached figure (Source)

However, in my runs, the velocity profile will be logarithmic in the end, and does not show the dip - even though it was clearly shown through a lab experiment that it should be there (actually quite pronounced).

I'm running this in interFoam, case setup is based on the Water Channel tutorial. I've tried the following modifications:
- Turbulence Models: no turbulence model / k-omega / k-epsilon / LRR
- Boundary Conditions: slip / noslip
- Meshing: Mesh is a simple blockMesh. Tried different resolutions (meshes with 5mm cells up to meshes with 1cm cells / refinement towards walls with bi-directional grading)

Does anyone here have experience with this? Any hints on possible settings to get the correct profile?

Is there such a thing as a "too fine" mesh? (looking at the 1mm-variant).

[shock77]

Hi,


could you maybe post your numerical results? Do you have a 3D mesh?


In generell, I only know one case, when a mesh can be too fine. Its when you use a RANS turbulence model and your mesh is so fine, that you model your turbulence instead of solving it directly, although your mesh is fine enaugh to do a LES or even DNS. You have also to be careful with the logarithmic law because for low values of y+ the stress at the wall becomes negative and you create a backflow.


But in general noone would create such a mesh, because the calculation time would unnecessarily high.