CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   ANSYS Meshing & Geometry (http://www.cfd-online.com/Forums/ansys-meshing/)
-   -   periodicity in 2 directions (http://www.cfd-online.com/Forums/ansys-meshing/93340-periodicity-2-directions.html)

s_azizmohammadi October 12, 2011 06:31

periodicity in 2 directions
 
Hi everyone,

I want to create a unique mesh with periodicity in 2 directions simultaneously, actually in 2D, :D. I know that it doesn't have physical meaning but I need this feature to check the capability of a method that I will explain later.

The problem is, I have a rectangular geometry, after meshing I want to populate some properties into the elements. Then I want to test it under 2 different situations (BCs), each one has a periodicity in one direction only. But I don't want to change the mesh configurations (element size, no of elements, ...) which will occur due to periodicity direction because I have to preserve my properties map all the time.

Any suggestion would be appreciated in advance.

PSYMN October 13, 2011 16:45

Which Mesher/Method are you using...?

s_azizmohammadi October 14, 2011 14:00

ICEM mesher 13.0

PSYMN October 14, 2011 17:00

By default, ICEM CFD doesn't support Dual Periodicity (or matching as you would see in ANSYS Meshing).

However, you can still fake it by copying edge parameters or curve node spacing from one edge or curve to another...

You said ICEM CFD, but not which method...

For hexa blocking, you can copy the edge parameters from one side to the other to get your periodicity.

For Patch Conforming, you can set the same curve distributions.

For unstructured 3D meshing (such as Octree Tetra), you will start with fixing the edges, but then may also need to copy shell mesh from one periodic side to the other, then merge the nodes with a tolerance and run the volume fill again... I am pretty sure I explained this in detail before... You can look it up here in CFD-Online...

s_azizmohammadi October 14, 2011 17:20

Thanks for your reply. Sorry I am beginner in ICEM CFD mesher.

Sure, I will look for your previous stuffs.

I want to do the periodicity in 2D not 3D and if I remember the method is called "2D planar", is it right?

I tried to resolve the problem by adding 4 extra fake nodes at the original corners of the rectangle, but it doesn't let to define these new points as a new periodic vertices, because in offset parameters we have to set the periodicity direction and if I put (1 0 0) it means there is only one periodicity in x-direction and if I set (1 1 0) it takes opposite diagonal vertices as periodic points.

Ok let me check your previous comments in the forum then I will be back.

PSYMN October 14, 2011 21:13

defacto dual periodicity
 
Ok, now that I know you are talking about 2D planar blocking...

I would suggest you create points along each side for roughly where you want the verts located. Copy the points to the other periodic side and then associate the vertices to the points to ensure they are periodic. Then for the rest of the edge distributions, you just copy from one side to the other...

You should end up with defacto dual periodicity.

s_azizmohammadi October 17, 2011 13:51

2 Attachment(s)
Are you sure this method guarantees periodicity in 2 direction simultaneously?

I already created 4 additional points in my model (as a new "Parts"). If I don't set any offset parameters when I want to pick the vertices I get this error: Attachment 9604

If I set the offset parameters just in one direction, when I want to pick the other points (that I virtually added to the model for the second direction) I get this error: Attachment 9605

I think the main problem is the offset parameters. How can define it when you want to have periodicity in both direction at the same time?

The other problem is when you create additional points it doesn't mean that these points belong to your geometry. Even if they are embedded in your geometry, when you want to mesh the geometry how can mesher distinguish original points and these fake points? where have to be meshed? I mean if you add extra points mesher conflicts to identify which geometry it should take into account. Isn't it?

PSYMN October 17, 2011 16:20

It is "defacto" periodic...

You won't be able to define those verts as periodic in Hexa because only one direction is supported, but if the vert locations and edge distributions match in the other periodic direction, then the periodic checks will pass when this is sent to the solver...

s_azizmohammadi October 18, 2011 04:22

Ok let say it is possible. What about offset parameters? Should I set (1 1 0)? or anything else?

s_azizmohammadi October 18, 2011 06:13

Now, I got the point :). I have the periodicity in both direction.

If I check the periodicity in "check mesh" and it says "no problem found", does it mean everything is ok?

PSYMN October 21, 2011 09:22

The check is just comparing the nodes in the parts you specify with the periodic setting... So it is really only checking in the one direction.

I guess maybe you could go into the periodic settings switch it to the other way and run your check again, but it may just be easier to send it to the solver and see if that complains...

Simon

s_azizmohammadi October 21, 2011 09:34

ok.

thanks a lot for your patient.

All the best.

s_azizmohammadi October 21, 2011 10:02

I checked it. Unfortunately :mad: it is not periodic in the other direction. So, it is still periodic just in one direction.

I am going to do the job from the beginning.


All times are GMT -4. The time now is 14:37.