[piyo]

Hi all,

I've tried using additional variable to model pollution dispersion. However, I cannot find a way to define or amend Turblent Schmidt Number for Additional Variable, does anybody know how to set it no matter in GUI or CCL?
I searched in the forum and noticed the new feature available in CFX13, but my version is 12.1. I don't think the number is the same as any of 'socalled' Turbulent Schmidt Numbers in the k-e model parameter tab, since these two constants are for flow field appearing in k and e equations.

Thanks for your help.
Piyo

[ghorrocks]

It will be in there somewhere as it is a fundamental constant for the turbulence model. But I have no idea where for old versions of CFX. You really should upgrade to the current version.

[piyo]

I set the number in Release 13's GUI and extracted the domain setting CCL. It doesn't cause any errors so far in CFX-PRE, but I wonder whether it really changes the constant in tranport equation.
I will test it later. I am quite sure this number cannot be amended in Release 12 or earlier versions, at least in GUI.

[Mina_Shahi]

Originally Posted by ghorrocks

It will be in there somewhere as it is a fundamental constant for the turbulence model. But I have no idea where for old versions of CFX. You really should upgrade to the current version.

Hi Glenn

is Turbulent Schmidt number always taken as constant in cfx (0.9)? or it can be changed automatically according to the type of flow or problem ?

actually i have two different types of mesh: in one i have better mixing (lower concentration gradient) while it has lower turbulent eddy viscosity compare to the other mesh. that doesn't make any sense .
because in this case if the turbulent eddy viscosity is lower, for the constant Schmidt number the turbulent eddy diffusion should be also lower which makes mixing worse, while it is not the case for me.

I mean i would expect higher eddy viscosity for having better mixing , is that right?


Could the numerical diffusion be the reason for having better mixing?

i would appreciate your help..

[ghorrocks]

Yes, the turbulent schmidt number is constant. It can be changed, but it must be a constant value for a simulation (no time or space variations) - at least as far as I can recall.

Unless you are an expert in turbulence models I strongly recommend against changing these constants. If you are having errors in your simulation a simple change to a parameter is unlikely to make things better. The values used are well established over a wide range of flows.

So if you have errors then do a standard CFD error analysis - see FAQ: http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

[Mina_Shahi]

Originally Posted by ghorrocks

Yes, the turbulent schmidt number is constant. It can be changed, but it must be a constant value for a simulation (no time or space variations) - at least as far as I can recall.

Unless you are an expert in turbulence models I strongly recommend against changing these constants. If you are having errors in your simulation a simple change to a parameter is unlikely to make things better. The values used are well established over a wide range of flows.

So if you have errors then do a standard CFD error analysis - see FAQ: http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

thank you for the answer, OK then turbulent Schmidt number is constant. fortunately i kept it constant (default value of 0.9) .

Originally Posted by ghorrocks

So if you have errors then do a standard CFD error analysis - see FAQ: http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

i don't know if that is an error, i am just comparing two similar cases with different grids (structured - unstructured). so i saw the behavior as i explained before .
in unstructured grid i saw less eddy viscosity and therefore less eddy diffusion while the mixing is better (higher diffusion rate and less concentration gradient) . so i was wondering if the numerical is the reason for that ! what do you think?

[ghorrocks]

If your mesh is too coarse for an accurate solution then you will get inaccurate answers. You do not learn much by comparing two inaccurate solutions. To compare hex versus tet properly you have to refine the meshes such that you have an accurate solution on both meshes to compare.

[Mina_Shahi]

Originally Posted by ghorrocks

If your mesh is too coarse for an accurate solution then you will get inaccurate answers. You do not learn much by comparing two inaccurate solutions. To compare hex versus tet properly you have to refine the meshes such that you have an accurate solution on both meshes to compare.

Thanks for replying me !
actually i compared the result of both grids with experimental data (just velocity). both seems promising however in the structured grid i had better prediction.
but for the mixing results is not what i expected. so i am thinking that numerical diffusion may be leads to higher mixing in unstructured grid. is that right?

[ghorrocks]

In some cases you will get higher diffusion in tet grids compared to hex, but CFX has done a pretty good job of making the diffusion on an equal quality hex and tet grid reasonably similar.

Other important factors for numerical diffusion are mesh size and mesh quality. Are the mesh elements the same size? Are they equivalent quality?