[Gearb0x]

Hello everybody

I'm currently trying to get a simple case working before attacking something more complicated. I'm using the simpleFoam solver since my flow is turbulent and incompressible.

The problem consists in evaluating the pressure drop between the inlet and the outlet of a canal.
I set up the boundary conditions like this :
- for p : I fix the inlet and outlet to p_inlet = X and p_outlet = 0 and see if the mass flow matches the one I have to have in my actual canal.
- for u : just fix it to zero @walls

Now, for the turbulent parameters k,nuT, epsilon, ... I test two cases : One with wallfunction and one with zerogradient @walls and no wallfunction. Normally, my mesh is fine enough to capture the details of the flow, I checked with the utilities "yPlusRAS" from openFoam, see the results in the pictures.

But I have quite huge differences between the simulation with the wallfunctions and without the wallfuctions and I don't understand why so much differences arise between the two ...

Do you have an idea where I could be wrong?

The results for P, U and yPlus :
- For the fine mesh, without wallfunction
P : http://yfrog.com/j9champpp
U : http://yfrog.com/j9champup

- for the same mesh, with wallfunction
P : http://yfrog.com/j9champpfp
U : http://yfrog.com/07champup
yPlus : http://yfrog.com/0ryplusp

[bastil]

Originally Posted by Gearb0x

Now, for the turbulent parameters k,nuT, epsilon, ... I test two cases : One with wallfunction and one with zerogradient @walls and no wallfunction. Normally, my mesh is fine enough to capture the details of the flow, I checked with the utilities "yPlusRAS" from openFoam, see the results in the pictures.

This is not really right. You either have a high Re-Mesh (y+ > 30 everywhere). In that case apply wallfunctions + an appropriate turbulence model. Otherwise you need a low-Re Mesh (y+ < 1). In that case you need a LOW-RE turbulence model. It is wrong setup to simply run with a high Re Model!
Additionally, you can not compare results from a low-Re mesh run without wall functions (which you seem to have) with the same mesh run wioth wall functions. It is wrong to run a low-Re-Mesh with wall functions!
Another topic is which wall function to use in 1.6 for low-Re models. It should be calculated for nuT afaik.
Hope this makes it clear.

Regards BastiL

[Gearb0x]

Firstly, a big thanks for your very clear and helpfull answer! I will now look for the LaunderSharmaKE turbulence model but I can't find a tutorial case wich uses this model.

There is just one thing I don't understand :
Comparing a low-Re mesh without wall function with High-Re mesh with wall function should be ok... no ?
Results should be slightrly different but not totally, am I wrong?

Purpose of wall function is to save computational time and to approximate the reality wich is represented here by the low-Re situation, no?

Thanks again for the help!

[bastil]

Originally Posted by Gearb0x

There is just one thing I don't understand :
Comparing a low-Re mesh without wall function with High-Re mesh with wall function should be ok... no ?
Results should be slightrly different but not totally, am I wrong

That is right. However, this means you have to create two meshes and not run both on one mesh.

Originally Posted by Gearb0x

Purpose of wall function is to save computational time and to approximate the reality wich is represented here by the low-Re situation, no?

In general, yes. However, low-Re remains a model and is also not "reality" but has fewer modeling assumptions.

[Gearb0x]

ok! I got it!

Thanks for the fast answer! It helps a lot