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/)
-   -   [ICEM] Shell X has node Y which has no twin (http://www.cfd-online.com/Forums/ansys-meshing/96377-shell-x-has-node-y-has-no-twin.html)

robboflea January 20, 2012 12:51

Shell X has node Y which has no twin
 
Dear all,

mesh check is failing on periodicity check. I get many errors of the type:

"shell 268991 has node 98118 which has no twin"

The translational periodicity in my model has been set-up correctly and all periodic vertices has been associated in the right way.

Can somebody please tell me what else could the problem be?

Thanks a lot

Rob

PSYMN January 21, 2012 19:48

Do they look periodic?

The periodic check gets pretty "technical" on you. It asks you what parts to check periodicity for and then reports exactly about those two parts. So if you have a curve on one side that happens to be in a different part (lets say it is in "CURVES" and you ask it for a periodicity check between PER1 and PER2, it might start to complain because technically it can't find a perfect match in PER1 for the line elements along the edge of PER2 because they are sitting over in CURVES which it didn't check... At least that usually turns out to be the problem for me.


If you don't think that is the problem, you can use subsets to help you get more out of that message. It tells you a node number, so go into mesh subsets (its a branch of the tree) and create a new subset by node number... Add a few layers and then compare that with what you see on the other side. Do they look ok? Can you see an obvious difference?

robboflea January 22, 2012 14:24

Dear Simon,

thanks for the help. I had all the periodic surfaces on 2 parts: "periodic1" for one wall and "periodic2" for its correspondent. The curves are in the "geom" part, together with the points. Do you think this could be causing the problem?

The mesh looks periodic. Also it has been obtained from a starting 2D mesh that has been extruded. The original 2D mesh is periodic and does not fail the check (and has been also successfully imported and analyzed in fluent).
I tried to take a look at the nodes where it happens by creating subsets as you suggested but I didn't notice anything irregular to be fair.

Rob

PSYMN January 24, 2012 00:08

Yea, the problem may be related to the curves being in the geom part... When you checked the nodes with the subset was it nodes along those curves?

I guess it doesn't hurt to put the curves in the per1 or per2 parts and try again...

Or if you think everything looks good, you could just send to the solver see how that likes it.

robboflea January 30, 2012 14:12

Dear Simon,

sorry for the late reply but I could not work on the problem the last week.
I tried to import the mesh in fluent but it crashes during import.

Have you got any other suggestions?

Rob

PSYMN January 30, 2012 18:21

No sorry, I would need more info to generate more suggestions...

robboflea January 31, 2012 09:48

1 Attachment(s)
Dear Simon,

I did a more deep investigation of what happens by creating subsets and I noticed that, apparently all the errors are coming from only two blocks (see the picture attached).
What I don't understand is what's happening. The periodicity is assigned correctly to the vertices of the blocks and so it's the same for edges association.

What can it be?
Thanks again!

Rob

robboflea January 31, 2012 09:52

Ah, I forgot to tell that the subsets in the picture are not the only ones giving error but they came from a random sampling of all the nodes in the error list.
I think that all the nodes in the periodic surfaces of those two blocks are source of periodicity errors but they are the only ones in the mesh.

Rob

robboflea January 31, 2012 12:50

Keeping investigating I noticed that (referring only to the two blocks giving periodicity problems):

1) all the elements on the periodic boundaries are interested by the error but not all the nodes on the periodic boundaries are
2) by decreasing the number of nodes on the periodic edges (I tried the value of 5 elements) the error disappears.

Rob

robboflea February 2, 2012 09:36

I found a solution to the problem. I don't know what ICEM was doing but I guess that extruding the original mesh gave it some kind of problems of unknown nature.

I tried to delete one of the two blocks and then recreate it again, reassigning the periodicity to the edges and it worked.

Hope this will help whoever will have the same problem in the future.

Rob

PSYMN February 2, 2012 11:00

So we never figured out what was actually wrong, but at least you learned how to use the subsets to locate the problem and then you were able to figure out a way around it... Good job.

jensi_t April 14, 2014 08:22

Same Problem, same Solution :).
Thank you


All times are GMT -4. The time now is 06:06.