CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   What is empty boundary condition? (http://www.cfd-online.com/Forums/openfoam-solving/64298-what-empty-boundary-condition.html)

lakeat May 6, 2009 10:53

What is empty boundary condition?
 
What on earth does "empty" mean in OpenFOAM?
  1. Say I have a 2D problem (usally N_z = 1), front and back bc are set to empty, is it the same to set them as cyclic? How does OpenFOAM treat "empty" bc?
  2. A N_x \times N_y \times N_z mesh, before mapFields, in order to save the time, I set N_z = 1, the bc in Z-direction that is front and back are set as cyclic, is it right in doing this?

henrik May 6, 2009 13:54

Daniel,

OpenFOAM does not have special machinery do deal with 2D domains. Therefore, a 2D mesh has twice as many faces on the front and back as it has cells.

Therefore it is a good idea to do nothing for those faces and that is what "empty" does.

Henrik

lakeat May 6, 2009 20:26

Then what if I set them as cyclic boundary, will they have the same effect? Sorry I cannot grasp the machineries, what on earth happens on each cell when Navier-Stokes Eq is solved?

(This is what I am thinking, if cyclic means "Enables two patches to be treated as if they are physically connected", then it (cyclic) is totally meaningless in a 2D case (N_z = 1), with no effect at all, so it would behave just as what empty bc does.) Please correct me if I was wrong. Thank you.

henrik May 7, 2009 04:13

Daniel,

the internal workings are not easy to explain in this context and by far off-topic for this thread. I suggest you have a look into Hrvoje's thesis.

Physically there is no z-component in the equations of your 2D problem which brings us intuitively back to "nothing" and "empty".

If it runs, cyclics will probably give you the right answer.

Note that the above applies to 2D domain, not wedges (2.5D domain).

Henrik

lakeat May 7, 2009 04:41

Okay, I have found if cyclic bc is used, nonuniform values come out in that patch, how did they come if no z-component in the equations, no interfaces in z-direction. But it seems these values are very small.

Another concern is about mapFields, if I have a 2D result in hand with empty patch for frontAndBack, and I want to map it to a finer grid with gridding in z-direction, cyclic bc for frontAndBack, can I directly change the empty to cyclic and then do mapFields?
Anyway, I tried this, and everything is okay so far.

henrik May 7, 2009 05:05

Daniel,

I understand you want to map the results obtained on a coarse 2D grid onto a finer 2D grid.

mapFields will map the cell values from one grid to another. The empty patches remain empty - There are just more empty faces in the fine grid. So what exactly concerns you?

Henrik

lakeat May 7, 2009 05:16

No, I mean map from a coarse 2D grid result onto a finer 3D grid, from 2D-RANS to 3D-RANS to LES. Is the procedure okay??


All times are GMT -4. The time now is 22:09.