CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   Question about the way to calculate the maximum wall cell width (http://www.cfd-online.com/Forums/main/107525-question-about-way-calculate-maximum-wall-cell-width.html)

Anna Tian September 29, 2012 19:33

Question about the way to calculate the maximum wall cell width
 
Hi,

For CFD case simulation, we need a relatively small wall cell width to catch the near wall fluid behavior. I don't have any experience to my new simulation case. So I have no idea about how to choose that value. May I ask if there's a way to estimate the wall cell width shall I set in my simulation? (of course, I will do the grid independence study later. But need a good initial guess.) If yes, how?

Thank you very much!

ski October 1, 2012 12:37

http://www.cfd-online.com/Tools/yplus.php

You specify the fluid properties, characteristic velocity and length scales, and the desired y+ value, and it will give an estimate on where to place the near wall cell.

Anna Tian October 2, 2012 07:23

Quote:

Originally Posted by ski (Post 384401)
http://www.cfd-online.com/Tools/yplus.php

You specify the fluid properties, characteristic velocity and length scales, and the desired y+ value, and it will give an estimate on where to place the near wall cell.

Thanks. But how do I choose the desired y+ value? Shall I always set it to 1?

SergeAS October 2, 2012 07:55

Quote:

Originally Posted by Anna Tian (Post 384499)
Thanks. But how do I choose the desired y+ value? Shall I always set it to 1?

It depends on the nature of your problem (such as tasks related to the heat transfer at the wall require small values ​​of y+) and the chosen turbulence model (each turbulence model has its own range of y+ where it is applicable)

Anna Tian October 3, 2012 07:57

Quote:

Originally Posted by SergeAS (Post 384505)
It depends on the nature of your problem (such as tasks related to the heat transfer at the wall require small values ​​of y+) and the chosen turbulence model (each turbulence model has its own range of y+ where it is applicable)

Thanks. I use K-Omega turbulent model. And I only care about the pressure drop. May I ask what y+ shall I choose? I'm still quite fresh to CFD. Is there any good tutorial on the choosing method of y+?

ski October 3, 2012 09:24

Ok, you'll need to work out if you're turbulence model is going to try to resolve the very near wall region, or is going to apply an empirical relation (a "wall function").

If you are not using a wall function, then y+<~1 is what you want. If you are using a wall function, then the very near wall region is not solved for (an emperical profile is applied instead). In this case, you'll need to check the documentation to see the y+ you'll need as it can differ depending on the type of wall function. Generally though, the 1st cell should be in the log law region at 30 < y+ < 300. Use y+=30 for best results. Some codes (most?) will give an option to use any y+ value by blending the wall function approach with a low Re model. This can be useful.

If you're not sure if you are using wall functions or not, then you'll need to check the documentation of the particular solver. If you have a choice, I'd recommend avoiding wall function unless CPU costs are a big concern.

lnk October 4, 2012 06:30

The CPU cost is always a big problem for me. May I ask what's the largest y+ could I use? If the first cell only need to be in the log law region, could I use like y+=290?

By the way, the first cell width shall also depend what value want to obtain, right? Like if for heat transfer, we will need much smaller wall cell width. I'm wondering is there also any theory about that dependency?


Quote:

Originally Posted by ski (Post 384721)
Ok, you'll need to work out if you're turbulence model is going to try to resolve the very near wall region, or is going to apply an empirical relation (a "wall function").

If you are not using a wall function, then y+<~1 is what you want. If you are using a wall function, then the very near wall region is not solved for (an emperical profile is applied instead). In this case, you'll need to check the documentation to see the y+ you'll need as it can differ depending on the type of wall function. Generally though, the 1st cell should be in the log law region at 30 < y+ < 300. Use y+=30 for best results. Some codes (most?) will give an option to use any y+ value by blending the wall function approach with a low Re model. This can be useful.

If you're not sure if you are using wall functions or not, then you'll need to check the documentation of the particular solver. If you have a choice, I'd recommend avoiding wall function unless CPU costs are a big concern.


ski October 4, 2012 06:45

Quote:

Originally Posted by lnk (Post 384865)
Thanks for your answer. The CPU cost is always a big problem for me. May I ask what's the largest y+ could I use? If the first cell only need to be in the log law region, could I use like y+=290?

By the way, the first cell width shall also depend what value want to obtain, right? Like if for heat transfer, we will need much smaller wall cell width. I'm wondering is there also any theory about that dependency?

In theory, your wall function will probably be applicable to such a large y+, but that dosen't mean that the results will be grid independent - they probably won't be. I'd recommend y+ around 30 if using wall functions (but check the documentation). Better still, resolve the whole boundary layer (without wall functions) and use y+ = 1. Either way, always do grid sensitivity tests!!

Wall functions are usually based on flow over a flat plate, so if your flow near the wall looks vastly different to this, then dont expect good results (e.g. for impingement). In this case, you shouldn't use wall functions and expect good results.

lnk October 4, 2012 07:31

Thanks. I'll always do the grid independence study. But I'd like to have a good starting point firstly. Some parts of my geometry are perfectly flat. I'll try quite large y+ there. Thanks for your hint.

By the way, if I more care about heat transfer, shall I use a smaller y+ than the case I only care about pressure? How small shall y+ in that case be?


Quote:

Originally Posted by ski (Post 384873)
In theory, your wall function will probably be applicable to such a large y+, but that dosen't mean that the results will be grid independent - they probably won't be. I'd recommend y+ around 30 if using wall functions (but check the documentation). Better still, resolve the whole boundary layer (without wall functions) and use y+ = 1. Either way, always do grid sensitivity tests!!

Wall functions are usually based on flow over a flat plate, so if your flow near the wall looks vastly different to this, then dont expect good results (e.g. for impingement). In this case, you shouldn't use wall functions and expect good results.


Anna Tian October 23, 2012 18:05

Quote:

Originally Posted by ski (Post 384401)
http://www.cfd-online.com/Tools/yplus.php

You specify the fluid properties, characteristic velocity and length scales, and the desired y+ value, and it will give an estimate on where to place the near wall cell.

Is the Re calculated here http://www.cfd-online.com/Tools/yplus.php the same as the Re we use to detect the flow regime?

What's L_{boundarylayer} in that formula to calculate Re? Is that characteristic length? Boundary layer length? Sometimes characteristic length can be quite different from the boundary layer length.


All times are GMT -4. The time now is 09:37.