CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Yplus (http://www.cfd-online.com/Forums/cfx/20216-yplus.html)

mefpz March 4, 2004 19:57

Yplus
 
Does Yplus have a reasonable range for the first point near the wall in CFX5.6?

Glenn Horrocks March 6, 2004 00:11

Re: Yplus
 
Hi mefpz,

If you are using the default k-epsilon turbulence model with default wall functions there is no reasonable range, it can handle pretty much any range of y+, so related issues such as mesh independance become the issue instead.

If you change the default wall functions to standard wall functions you will need to keep y+ above about 11. If you change the default k-eps turbulence model to one of the low-Reynolds number models you will need to use y+ less than 1.

Regards, Glenn

mefpz March 6, 2004 20:37

Re: Yplus
 
Thank you very much! I have read your answers to other fellows and I have got some pictures about yplus. Still I have some questions: 1: Would you like to specify some low-Renolds models? 2: When I use SST and LES, how to decide the yplus? Do they have ranges? Thanks in advance.

Glenn Horrocks March 7, 2004 16:53

Re: Yplus
 
Hi mefpz,

Have a look in the documentation for the low-Reynolds number turbulence models. They are the ones which don't use wall functions and require y+<1. I don't have the documentation in front of me at the moment so I can't give you the list.

The default wall function for the SST turbulence model does not have a y+ limitation.

I don't think LES is an officially supported feature yet. It might be a beta feature but I have never used it. In 5.7 a detached eddy simulation (DES) model will become available which looks like a useful combination of LES and RANS models. Have to wait a few months yet for that one to come out.

Glenn

mefpz March 7, 2004 22:46

Re: Yplus
 
For now I have no question on Yplus. Best Regards!

mefpz March 8, 2004 05:14

Re: Yplus,another question!
 
In your earlier answers you mentioned the low-Renolds models. I read the turbulence models in CFX5.6 documents, but found no useful information. I wounded whether the low-Renolds models are turbulence models(such as k-w model) that they are used in low-Renolds number or the ones(different from k-w model) Based on low-Renolds number?

Glenn Horrocks March 9, 2004 18:28

Re: Yplus,another question!
 
Hi mefpz,

I am no turbulence model expert, but I will attempt to answer your question.

There are low-Reynolds number turbulence models around, which, as their names suggest means they are designed to model turbulence at low Reynolds numbers. At high Reynolds numbers the boundary layer is fairly simple and reliable, hence wall functions are reasonably successful.

As the boundary layers of low Reynolds number flows are a lot less predictable and mean that simplifying it down to a simple wall function does not work very well. Hence low Reynolds number turb mobels are generally "integrated to the wall", right to the point where the flow is essentially simple shear (that is, right down to y+<1).

It should be remembered that although the turbulence model and the way walls are handled are linked, there are some options here. To model a low Reynolds number flow you can only choose a low Reynolds number turb model and integrate to the wall, but to model a high Reynolds number flow, you should choose a normal high Reynolds number turb model, but can use wall functions of integrate to the wall (ie no wall functions).

If you are going to intergrate to the wall in a high Reynolds number flow, you need a very fine wall mesh (you need to resolve y+<1), and you need a turbulence model which is defined at the wall. The turb models using a normal epsilon equation are not defined at the walls, so you cannot use these types of turbulence models. The omega equation turb models are defined at the walls, so they are a better choice if this is what you want to do.

Nothing in CFD is simple.

Glenn

Neale March 10, 2004 20:49

Re: Yplus,another question!
 
Well, you could just use SST and call it done. It applies k-omega near the wall and k-epsilon in the free stream, which is more appropriate there (k-omega is defficient in free stream flows).

The SST model in CFX-5.6 will automatically integrate to the wall where it can (i.e., if you have enough grid) and use scalable wall functions otherwise.

The point is, if you don't have enough grid to resolve your boundary layer then it wont try to resolve it by integrating to the wall.

Neale.


All times are GMT -4. The time now is 07:31.