I was looking at ICEM CFD's web site today and they claim they can automatically convert a tet mesh in to a hex mesh. This sounds like a very impressive tool, has any body got experience of using this?

6 tets can be combined to form a hex. so if you have a tet mesh you can easily build hybrid meshes with hexes and pyramids by combining the tets

TANSTAAFL

A tetrahedron may be subdivided into 4 hexahedrons, but you have to be willing to accept the level of distortion, which typically will degrade the quality fo solutions computed on the hexahedral mesh.

But, if you must have hexes, this drawback may be a better option than none.

Hello all,

The ICEM method of turning TETS in HEX uses a cluster of 12 TETS, the resulting HEX cells are near perfect cubes with pryamids to connect to the existing mesh. Overall you can reduce your cell count by 500% on average. If any body wants to see an example visit.

http://www.icemcfd.co.uk

Then click on the what's new box. This has a very simple example to show the theory but the complexity of the original TET model doesn't matter.

This type of mesh is suitable for most unstructured solvers such as CFX, STAR and FLUENT. If any body has any more questions please let me know.

Cheers

David

David,

What about the fact that the tet mesh must be unsmoothed before you do the tet-hex conversion? I've tested the feature just briefly, but I found that this constraint leads to rather poor tet and pyramid quality near the boundaries. Also, any prism layers need to be generated before the conversion, and this process is affected by the need to have an unsmoothed tet "core" as well. Any comments will be highly appreciated.

Regards,

Lars Ola

Hello,

What version of ICEM are you using? Version 4.1.3 is the most upto date and with this it doesn't matter if the TET mesh is smoothed or unsmoothed prior to the conversion. I would expect the following quality values for the cells:

TETS 0.08 to 0.15 aspect ratio HEXAS near 1 determinant PYRAMIDS 0.03 to 0.1 aspect ratio PRISMS 0.05 to 0.1 determinant

These numbers are a bit lower than a standard TET / PRISM mesh but we have UK users running these meshes is both Fluent and STAR CD with out any problems. Inital cases show that the automatic HEX mesh gives a more accurate solution at a faster solution time.

I hope this helps?

Regards

David

David,

I've tested the tet-hex conversion feature on v4.1.1 only, but we're in the process of installing v4.1.2, which we recently received from Computational Dynamics upon request. CD claims that they send out the CD-ROMs of new versions as soon as they receive them, but it seems impossible to stay even remotely update unless downloading the new versions directly from ICEM's webpages every other week. This is really not satisfactory at all considered that the users need to focus their efforts on their daily work. In my opinion, immediate notification should be sent out to all licensees upon the release of important bug fix patches, and major releases ought to be dispatched to the customers without prior request.

But that discussion probably belongs in a different forum. The tet-hex conversion capabilities of v4.1.3 do appear to be quite impressive, and I will test them as soon as I get the opportunity. Meanwhile, how does it perform in v4.1.2?

Regards,

Lars Ola

Hello,

I cannot comment for CD but the policy at ICEM CFD UK is to send all new versions automatically to users as soon as they are released. Please request that CD do this for you or contact your local ICEM office and ask to be added to their distribution list.

4.1.2 tets to hex on some examples may give uncovered faces. You can get about 2 cells per million which will have to be fixed by hand. 4.1.3 with all my cases so far doesn't have this problem. If you run by command line you also have the option to set minimum cell quality. This is useful if your solver is more sensitive to cell shape, this will mean you get fewer HEX cells in the final mesh.

Regards

David

David,

I also experienced uncovered faces in v4.1.1, but they were a lot more than two, and the model didn't even contain half a million tets. Also, ICEM crashed when trying to smooth the converted mesh. Are there any special precautions to be made before smoothing in v4.1.2, and can an already smoothed mesh be converted in v4.1.2?

Regards,

Lars Ola

Hi,

Yes you can use any mesh, smoothed or unsmoothed in 4.1.2 but watch out for the uncovered faces. If you want to use this feature it really is worth getting 4.1.3. If you are in Europe please call

004951128869653 and ask for a CD to be sent to you.

Regards

David

Hi David,

I think I will. Does "Europe" mean all of Europe or just continental Europe? Btw: Our direct CAD interface to I-DEAS (v4.1) hasn't been updated since we changed to v4.1.1. Is there perhaps a newer version for that too?

Regards,

Lars Ola