[R4z0rX]

Hey guys,

I am modeling a flow over pretty rough walls with equivalent sand grain roughness of 73.29 µm.

I read the solver theory guide about the rough wall treatment and found out that in fully rough flow (hs>=70) the viscous sublayer breaks up completly and thus can be neglected.

How does CFX handle the rough wall in this case in combination with the SST turbulence model? Can the automatic wall treatment still be used or what adaptions must be made? I read something about that the max yplus must be below 11.06. I assume that the first layer thickness should not be lower than the sand grain roughness. Is that correct?

Hope you can help me a little bit with this issue.

Thanks alot!

[ghorrocks]

A key point is that if you have a very rough surface then very high mesh resolution at the boundary layer makes no sense. A guide is that if the sand grain size is more than an element length or two then you need to think about whether you have too fine a mesh.

So for rough surfaces you should not be integrating to the wall. You should be using wall functions with the first element significantly off the wall (including sand grain roughness) such that the first node is in a region where a traditional boundary layer is formed, not the disturbed flow around each individual sand grain.

[R4z0rX]

Thank you very much for your response!

So the theory guide states that the wall is placed at 50% of the sand grain height.
So I have to enter a first cell height. Should I enter 0.1 mm as first cell height, since that is a little over 0.07329 mm which is the sand grain roughness or should I enter 0.04 mm as first cell height, since that is a little over half the sand grain roughness?

I guess this is dependent on where CFX places the actual wall. Do you know which one is the right approach?

Thank you very much!

[ghorrocks]

As the documentation states, the wall is placed at the 50% sand grain height. So that is where the wall is.

You should determine your near wall mesh resolution by a mesh sensitivity study. Different simulations have different sensitivities to this, so you need to check for what you need in your case.

[R4z0rX]

Thanks for your help!

I have read the technical report from Lechner and Menter "Development of a rough wall boundary condition for w-based turbulence models", which suggests to use scalable wall functions for rough walls especially in fully rough flow (hs > 70), so that the min yplus value can be set to 2.5 instead of the standard value of 11.06.

Besides that it suggests some corrections for the heat transfer.

This option is not available in the CFX-Pre user interface, so it has to be changed using CCL-files.

Does the CCL-file has to contain the complete setup or can I add only certain things I want to change so that the solver only uses these corrections (and overwrites thes settings if necessary) from the CCL and leaves all the others as defined in the user interface?

[ghorrocks]

You can define a CCL which is just your changes and everything else will be left unchanged.

The way I do it is to use the command line "cfx5cmds -read -text RunName-001.ccl -def RunName.def" to generate a full CCL for that run which you can then edit and write back with "cfx5cmds -write -text RunName-001.ccl -def RunName.def". This also keeps a handy reference of what your ccl was for that run in the same file structure as all the other files.