k-epsilon, enhanced wall treatment
 

Hi,
I have compared CFD results (Fluent 12) and theoretical results (by dimensionless numbers) at very simply geometry - straight pipe. I compared the pressure losses and change of fluid temperature in the section of pipe.
I used a few meshes with diffrent quality (y+ = 0.5 to 300).
I set up the model k-epsilon RNG, Enhanced wall treatment.
The Fluent help say that the Enhanced wall treatment should be used only in case very fine mesh (y+ is aproximately 1).
So, I was very suprised, that the CFD results corespond to theory very well - in case of very fine mesh and also in case of very coarse mesh (y+ = 300). The results look realy independent on mesh quality and y+. I hope that the good results are not only a fluke.

My questions: Do you thing that is posible to use Enhanced wall treatment for coarse meshes with big values of y+? Can I supose relevant results in case I solve more complicated problems with locations of small y+ values and also with locations with big y+ values in one mesh?

Thank you for your opinions and sorry for my bad English.

Hey Michal,

If u r using SA model, u either use mesh with y+ between 30 and 300 or y+ around 1, i that u dont need to use enhanced wall treatment or wall funtion.

Similarly for k omega and its variants, u follow the rules for wall function, but if u activate transitional flow in the define>model, u have to consider rules as for enhanced wall tratment.

For k epsilon, u have to use wall function if ur using coarse mesh. See the Fluent user manual, they have explained very clearly with figures the difference between enhanced wall treatment and wall function.

Hope it helps

Hi Mariachi,
so, it is posible to use SA model in wide range of y+ ? (aprox 1 to 300 ?) .
Yes, I have read manual - I have been finding acceptable turbulence model which is usable for mesh with both - small and big - values of y+. I need to solve flow and heat transfer in synchronous generetor and it is very dificult to make complicated mesh with good values of y+. Unfortunately, one posible way for me is upgrading the mesh at location of non-OK y+. So thank you for our response.

Best regards
Michal

Michal,

SA model will give u best results if

y+ values between 30 and 300 (better if they are as close as possible to 30)
OR
y+ values <=1

In your case i suppose u have huge variation in y+ values (between 1 and 300) right? If that is so, then using SA model will give u best result at the places where y+ is either less than 1, or y+ is between 30 to 300. So at all other places where y+ is between 1 and 30, numerical errors will occur which may result in corrupting ur solution significantly (depending on what level of accuracy ur trying to achieve).

I know, just meshing a complicated geometry is quite hectic and frustrating, and if ur trying to make good quality mesh on a complicated geometry, is a nightmare. Try to find out some way to make good quality mesh because if ur mesh is good, it will eliminate most of the problems in Fluent that u may encounter otherwise, and ull get quality results hopefully.

So my recomendation is to make a mesh such that either all y+ values are less than 1 or between 30 and 300.

The reason why we shouldnt have y+ between 1 and 30 lies in the background equations and assymptions behind SA model. Read Fluent user manual if ur more interested and ull understand.

Goodluck

Hello everyone,

I've been working on CFD too. I would learn about "y+", "wall function" as long as "wall treatment" etc. I known every body known how to estimate the y+ value, so can you show me how to do it (in very detail if posible :) ). Further, can you explain what is the "wall function", "wall treatment". So far, I want to discuss about another things, that is CFL number and time-step.
I'm very glad to discuss with everyone! :)
Email: trieuckgt@gmail.com
MSN:trieu.dut@live.com or YM: trieu_tme@yahoo.com
thanks for all!