[MissCFD]

Hello,

I would like to know if the fact of having y+<30 in some places (where we don't have many controls on the strutured mesh) may affect calculation when we use k-epsilon turbulence model ? Apart from these places, the values of y + is between 50 and 130 and the overall average is y+=50 for all the mesh.

Thank you for your help/advice

[ghorrocks]

There is no universal answer to that. If there was it would make sure it always happened. Some simulations are sensitive to it and some are not. The best way to be sure is to generate a few meshes with different y+ and see if it makes a difference in your case.

[LorenzoSaintDubois]

Shouldn't the scalable wall function take care of it if the grid gets too fine close to the wall?

[ghorrocks]

No! Don't get confused by scaleable wall functions. The y+ range she mentions is way above where the scalable wall functions will do much. Scalable wall functions only do anything when y+ goes below 11. In the regime she is working in it will be using standard wall functions and so the answer to her question is mesh size sensitivity.

But as I said, there are some flows where resolution of the boundary layer is not too important. If this is the case then you can mesh it with any y+ you like and it will not make a difference.

[MissCFD]

Thank you very much for your answer ghorrocks.

But what happens when the k -epsilon model is used in a place where the y+ value is less than + 30 ? The result is not valid for this place ?

I have read in an other post that for the k-epsilon model, we must have y+ value between 30 and 50. Is it fair ? Because I have read in CFX user's guide that to use standard scalables functions, we must get y+ > 30 and y+ <= 300. So, to get y+ value > 30 almost everywhere in my domain, my y+ values can reach 150. Is it right ?

Originally Posted by ghorrocks

There is no universal answer to that. If there was it would make sure it always happened. Some simulations are sensitive to it and some are not. The best way to be sure is to generate a few meshes with different y+ and see if it makes a difference in your case.

[ghorrocks]

Most people say the normal wall function approach is OK down to 11. So you should be OK at 30. And the scalable wall function approach allows you to go below 11 without problems (usually).

y+ is going to vary as the flow speed varies, and you mesh bends around things. You are always going to get a big range in y+ values. But before you do lots of work on y+ you should work out if your simulation is sensitive to y+ or not. If your simulation is not sensitive to y+ then it will not help.

There is no fixed upper limit to allowable y+. It is set by mesh sensitivity and resolution of the boundary layer - and that varies for each simulation.

And finally, there may be regions in the simulation where the boundary layer is important and areas where the boundary layer is not important. If y+ is high or low in a region which is not important then it does not matter, does it?

[MissCFD]

In CFX 14.5, when we use the k-epsilon model, we don't have choice, it's only the scalable function that is proposed. So, if I understand correctly, the scalable wall function used with k-epsilon model don't use nodes of my mesh having y+ <11.06 ?

For example, if at a place in my mesh the first node give a y+ value of 3 while if we consider the second node we reach a y+ value of 15, it's the second node only that will be taken into account ? Is that right ?

It's not very clear for me, I wonder if below Y+ = 11.06 mesh points are ignored or below Y+ = 11.06 mesh points are set up to y+ =11.06.
But if it's the last option, this means that we associate a y+ of 11.06 on cell initially too small, do you see what I mean ?

Originally Posted by ghorrocks

Most people say the normal wall function approach is OK down to 11. So you should be OK at 30. And the scalable wall function approach allows you to go below 11 without problems (usually).

y+ is going to vary as the flow speed varies, and you mesh bends around things. You are always going to get a big range in y+ values. But before you do lots of work on y+ you should work out if your simulation is sensitive to y+ or not. If your simulation is not sensitive to y+ then it will not help.

There is no fixed upper limit to allowable y+. It is set by mesh sensitivity and resolution of the boundary layer - and that varies for each simulation.

And finally, there may be regions in the simulation where the boundary layer is important and areas where the boundary layer is not important. If y+ is high or low in a region which is not important then it does not matter, does it?

[ghorrocks]

Not quite right. The scalable approach assumes that if on the first node y+<11 then it just makes it 11. This means it actually shifts the boundary layer position a little, in effect meaning the wall for the purposes of the boundary layer is a little further away than it actually is. It does not ignore nodes.

[MissCFD]

I'm sorry but I'm not sure I clearly understood the meaning of your sentence "This means it actually shifts the boundary layer position a little, in effect meaning the wall for the purposes of the boundary layer is a little further away than it actually is."

Also, you said "it just makes it 11", does that mean that the node of mesh is moved to reach this value ?

Originally Posted by ghorrocks

Not quite right. The scalable approach assumes that if on the first node y+<11 then it just makes it 11. This means it actually shifts the boundary layer position a little, in effect meaning the wall for the purposes of the boundary layer is a little further away than it actually is. It does not ignore nodes.

[ghorrocks]

No, the mesh is not moved. If y+<11 then you just apply the wall shear stress as if it was 11, that's all.