CFD Online Discussion Forums

CFD Online Discussion Forums (
-   ANSYS Meshing & Geometry (
-   -   [ICEM] periodic tolerance icem tetra / hexa (

Jonathan December 15, 2011 08:04

periodic tolerance icem tetra / hexa
Hi Everyone

just wondering whether anyone knows whether there is a tolerance associated with the periodicity function in ICEM.

i.e. if one specifies a rotational periodic angle of 18 deg in GLOBAL MESH SETUP, and the actual periodic faces vary from say 17.9 - 18.1 deg, will ICEM have a problem?

I am getting quite a few errors in tetra when trying to mesh a periodic flow volume in which i think the geometry angle between the periodic faces varies slightly ...

any ideas???

much appreciated

PSYMN December 19, 2011 12:22

I am not sure about what the tolerance is, but I would expect a problem if you said 18 and it varied by as much as 0.1.

Maybe varying by 0.001 would be fine.

What you should expect to see is the mesh trying to maintain the 18 degrees in spite of the geometry. Issues may result if you don't like its guess as to where that 18 degrees should fall.

Jonathan December 19, 2011 13:47

1 Attachment(s)
hi Simon,

thanks a lot for your reply,

Yes, i think this is what i ended up seeing. mesh elements running over the geometry curves / jumping from one face to another. i ended up finding a geometry fault where gambit was throwing away the first data point for a curve effectively shortening the periodic arc by a fraction of a degree, but obviously too much for ICEM, which was then having issues! incidently i probably wouldnt have found the mistake had i not tried to mesh in ICEM!

on the same problem, i have had a lot of trouble getting tetra to give me a workable model (i am trying to migrate my batch routine meshing off of gambit into tetra) and even though i am sure i have no more geometry related periodic issues, i am ending up with problems in almost every mesh approach i have tried so far unfortunately :eek:

the geo is the same turbine i looked at briefly with hexa but because of the need to come up with a robust meshing approach (the meshing routine is part of an bigger turbine optimisation routine), i have been looking at tetra + prisms rather than trying to script a hexa journal which i feel might not be fool proof as the turbine geometry changes slightly through each optimisation iteration ...

basically i am:
1)importing ACIS geometry
2)running build mesh topology
3)creating various parts (inlet / outlet / blade / casing etc)
4)defining the MAT point
5)Octree meshing to get the surface meshes
6)smoothing then remeshing the volume with Delaunay
7)running prism using '0' initial height (this function is not available in gambit, hence the motivation to switch)
8)remeshing with hexcore (seems to be better than the gambit hexcore routine) to get the final flow volume

in general the order in which ICEM meshes (surface / interior / prism / then hexcore) is better than gambit's which ends up trying to squeeze pyramids and tets between the hexcore & prisms at the end, giving you very poor quality tets / pyra in the final mesh!

however, so far with ICEM i have also ended up with a few issues which i was not expecting (unmatched periodic nodes / faces (no periodic twin), tri faces belonging to periodic part A on periodic face B (!) etc).

not sure if this is due to my setup (likely :D) or tolerances which need to be tweaked, but i think so far I have only managed to get x1 workable mesh out to Fluent so far!!

anyway, apologies for the long winded msg, any help would be greatly appreciated,

best regards

CONSOLE output:
Running diagnostics for Duplicate elements in subset "all"
No problems were found for Duplicate elements
Running diagnostics for Uncovered faces in subset "all"
No problem volume elements were found for Uncovered faces
Running diagnostics for Missing internal faces in subset "all"
No problems were found for Missing internal faces
Running diagnostics for Volume orientations in subset "all"
No problems were found for Volume orientations
Running diagnostics for Surface orientations in subset "all"
no orientation errors
faces are correctly oriented
Surface orientations are OK
Running diagnostics for Hanging elements in subset "all"
No problems were found for Hanging elements
Running diagnostics for Periodic problems in subset "all"
triangle 302 has 3 twin nodes but no twin face.
triangle 3518 has 3 twin nodes but no twin face.
triangle 16678 has 3 twin nodes but no twin face.
triangle 16735 has 3 twin nodes but no twin face.
maximum distance from periodic 3.4857e-005
between vertices 2757 and 46455; Fixed
there are problems with the periodicity

Running diagnostics for Multiple edges in subset "all"
No problems were found for Multiple edges
Running diagnostics for Triangle boxes in subset "all"
No problems were found for Triangle boxes
Running diagnostics for Single edges in subset "all"
No problems were found for Single edges
Running diagnostics for Non-manifold vertices in subset "all"
No problems were found for Non-manifold vertices
Running diagnostics for Unconnected vertices in subset "all"
3225 unconnected vertices were found.
Unconnected vertices are OK

Jonathan December 22, 2011 10:02

Hi Guys,

Just wondering whether anyone has any idea why this meshing approach / geometry will not make a acceptable periodic grid for my solver ...

I have read as many (pretty much all!) the threads on ICEM periodicity and find nothing which leads me to a solution ... I, at the moment, cant understand why a simple tetra mesher cannot build a periodic mesh on this geo?? it seems like a fairly simple task??

all (any) assistance would be greatly appreciated ...

many thanks ...
best regards

PSYMN December 22, 2011 12:26

Assuming you have now fixed the issues with the periodic geometry, I am not sure why you would be having issues. This is the kind of problem that we would really need to see to diagnose. You could send it in to Tech support at

Alternatively, your issue seems small, just a few triangles in the wrong part, so you could just fix them manually an move on.

Find the few triangles in the wrong part and add them to the right part. Then the periodic check will work.

All times are GMT -4. The time now is 11:17.