Pyramids on mesh

[Arnaud Barthet]

Hello,
I generate an unstructured mesh with icemcfd 12.1 with a mix of tetra/prism but in some regions of the mesh there are pyramids which are created. I usally convert my mesh to unv format but I cannot do this because of pyramids. How can I build a mesh readable by Code_saturne with these pyramids? I have seen that with cgns it can be done but I have had a crash when I have launched ecs convertor.
thanks for your help.
Arnaud

[Mickael Hassanaly]

Hello Arnaud,
 
Could you post your mesh file here? If not, send us your exact error message or try med format.
 
Best regards
 
Mickaël

[Yvan Fournier]

Hello Arnaud,

What version of the CGNS library did you build Code_Saturne with ? ICEM made a mess of CGNS export somewhere between version 11 and version 13, as they moved to CGNS 3 (beta), but not necessarily to 3.1. CGNS element numbering went through several changes between 2.5 and 3.1 (CGNS 3.0 was only released as a beta), and ICEM seems to have added to the problem by just setting a CGNS file version to 2.5 when export to CGNS 2.5 is selected, but using element type numbers from CGNS 3.0.

I am not quite sure, as the issues may be different with ICEM 12 and 13: in one case, building the Code_Saturne preprocessor with CGNS 3.1 support (instead of 2.5) solved the issue, while in another, We had to use the CGNStools cgnsview to edit the CGNS version of a file ("cheating" on the version number, or rather, compensating for ICEM's cheating on that number), so that it would be read correctly.

As ICEMCFD does not to my knowledge provide MED format export, and the Ideas universal file format is an obsolete piece of junk with no provisions for pyramids despite all the bloat added over the years, building Code_Saturne with CGNS 3.1 support (and experimenting with forcing the CGNS version number to 2.5, 3.0, or 3.1 in the CGNS file using cgnsview if this is not enough).

Best regards,

  Yvan

[James McNaughton]

I can't get the CGNS exported from ICEM to work in Saturne either. If you have Fluent you could try what I am currently doing:

• Export from ICEM as .msh (Star ccm+ format in the output).
  Importing the .msh to Fluent.
  Exporting as CGNS from Fluent (which works fine with Saturne).

Hope that helps if you can't get anything else to work!

Yvan, are there any plans for Saturne to be able to read .msh files in the near future?
 
Thanks,

James

[Yvan Fournier]

Hello James,

Several users have requested this feature, but none so far have been able to provide me with the documentation for the ANSYS msh format.

If somebody implements support of those files in Code_Saturne (and provides me with a link to the documentation of the format so I can check that a broad enough range of the format is supported), I'll integrate it and support it in the code, but otherwise, it should not be Code_Saturne's job to work around ICEM's bugs. I have heard from users that recent academic versions of ICEM do not include CGNS output at all, and that it requires an additional licence, so I understand the need for another format, but if ICEM is starting to practice user lock-in, chasing a potentially moving target but supporting their msh format may not be easy. If the format documentation is publicly available, things get better, but I have not been very lucky searching for it.

I have not tested it, but if you need a robust commercial solver,with CGNS output, Pointwise seems to be a powerful tool, and might be a better option than ICEM. Also, NUMECA's HEXPRESS tool seems able to output CGNS. Here again, we only tested an old precursor to that tool, which had its own format, but in 2001, it seemed quite promising, and it has had a few years to get better, so you may want to give it a try.

[James McNaughton]

Hi Yvan,

Thanks for the information about the meshers, I've heard good things about Pointwise, unfortunately I'm pretty much restricted by my University's choice of software (our gambit licence having recently run out hence the move to ICEM). I'm not restricted to CGNS, only what Saturne can read, my current work around seems to do the trick (for now!) so I am happy to stick with that.

I imagine that unfortunately the documentation you require might not be in the public domain. It is good to hear you would help with the integration so if I hear from anyone that they can help then I will be sure to investigate.

Thanks again,

James

[Arnaud Barthet]

Hello,
thanks for all your answer but if I well understand, if I use ICEM I cannot export a mesh with the good format if there is pyramids in my mesh? So the only way is to find the good parameters in order to generate a mesh without pyramids.
Thank you,
Arnaud

[Yvan Fournier]

Hello,

Apparently, users of ICEM 13 are able to generate valid CGNS meshes by exporting to "CGNS 2.5", then using the cgnsview utility from CGNS to edit the CGNS version number in the files.

I also just tried exporting to CGNS 3.0 using ICEMCFD 13.0, with the HDF5 variant of CGNS (not ADF), and boundary conditions on faces (not nodes), and was able to read it normally using the Code_Saturne preprocessor built with CGNS 3.1.3 support, so things seem to have improved.

So if you can upgrade from ICEM 12.1 to 13, your problem may be solved. Otherwise, you could try the StarCCM+ output in ICEM, but you then need to install libccmio (Google the "Mandriva Linux Code_Saturne" for installation instructions, in French), and build Code_Saturne using this library; so this is really a last resort (we have tested that format with files produced by StarCCM+, not by ICEM).

Best regards,

  Yvan

[Yvan Fournier]

As a follow-up, using ICEM 13 and exporting as CGNS 3.1 seems to solve the issue, by at least 2 accounts (including a test of my own).