Reading UNV mesh (SALOME) with Saturne.

[Marie Pomarède]

Hi there,
 
Maybe this message does not have to appear in this topic, please tell me if I have to create a new discussion or anything else.

I also have a problem with a .unv format mesh that I created with ICEM CFD, and exported for Code_Saturne v 1.3.3. All groups and colors are well defined if I compare them to what I created. But, I think that the fact that my mesh contains 'pyramid' elements, I mean five-faces volumes is a problem.

It seems that either the .unv format does not support such elements, or Code_Saturne does not recognizes them : I invariably get a message informing that some boudary faces have not been coloured and does not belong to any group, while all the boudary faces I really created are well defined and belong to right groups.

I quickly read IDEAS documentation and I didn't find any element that could correspond to my pyramids.

Would you have an idea about this ? Did someone encounter such a problem with those elements ?
 
Many thanks for your help,

Regards,
 
Marie

[Yvan Fournier]

Hello,

You are right, in that The unv format does not allow for pyramid elements. So when a meshing tool exports meshes to unv, it has 2 options:

discard pyramids (leaving holes in the mesh) write pyramids as hexahedra with 4 identical vertices (handled by Code_Saturne)
ICEM seems to choose the first option, which is why we recommend ICEM users to export in CGNS rather than in unv format. A new problem reported by some of our academic users is that that recent "academic" licences of ICEM do not seem to support CGNS export by default (this now requires an aditional module which costs extra). In this case, depending on the complexity of your mesh, you may be able to use another meshing tool.

It would be interesting for academic users of ICEM to tell us if they have access to Star-CCM+ or Exodus-II output formats, as those could be backup output formats when CGNS is not available (we do not currently support Exodus-II, but we might in the future).

Best regards,

  Yvan

[Marie Pomarède]

Hello Yvan,

Thank you for your confirmation. I tried to export my mesh in CGNS format (2.4), but it seems that this format (maybe a bit different because it is created by ICEM ?) is not supported by Code_Saturne... Running the pre-processor give me a crash. I'm pretty shure that the ICEM license I am using is not "academic" since it is allowed to export various CGNS format (from 2.1 to later).

If I try to keep a UNV format using the second option that you propose, I mean write pyramids as hexahedra with 4 identical vertices, how am I going to go about it ? Do I have to make this modification manually ?

Thanks a lot,

Best regards,
 
Marie

[Yvan Fournier]

Hello,

Was Code_Saturne built with CGNS support ? As CGNS support is optional, Code_Saturne may have been installed without it, in which case you need to install the CGNS library and headers (which may be available directly as a package in recent Linux distributions), and re-install the Code_Saturne preprocessor.

What does your error message say ? "Unknown mesh format", or "file not open" ? I have recently noticed that with the CGNS package from Ubuntu (and probably Debian), we can get a "file not open" error after the file is opened, and this is due to their CGNS build handling HDF5-based CGNS files and not ADF-based CGNS files (the previous default). Simply running the "adf2hdf" converter (included with the CGNS package) make the file correctly readable (and "hdf2adf" converts thigs back the other way if you need).

Finally, writing pyramids as hexahedra with 4 identical vertices in the UNV format is only an option if you may control how the mesh export is done: if ICEM does not provide this option, you would need access to its source code to to it (not possible), or to export to another format than handles pyramids and from which you could write a converter to UNV, which would be quite a task, much more so than installing CGNS and re-installing the preprocessor...

Best regards,

  Yvan

[Marie Pomarède]

Hello Yvan,

Yes, Code_Saturne was built with CGNS support, and the file was opened, but I got a message (french) : "pile d'appel" after the reading of dimension and vertices number. But, doesn't matter, I fixed my problem : I was working with ICEM CFD v.12, which was linked with CGNS library version 2.40. It seems that the problem was here, because using ICEM CFD v.11, which is linked with CGNS library version 2.42, gives good mesh reading. Everything is OK (boundary conditions, pyramids are supported, etc.) and my calculation is running normally.

Thank you for the time,

Best regards,
 
Marie

[David Monfort]

Hello Marie,

That could still be interesting for us to have your CGNS mesh (the one taht fails to be read by the preprocessor). Do you think you could send it to us?

Thanks!

David

[Marie Pomarède]

Hello David,
 
of course I can, I'm sorry I didn't have the reflex to attach my meshes. Here you can find two very simple meshes (each is representing two coaxial together hanged tubes made with different mesh techniques), respectively called pyrav11.cgns for the one obtained with ICEM v11 and pyrav12.cgns for the one obtained with ICEM v12.

Hope it will help,

Regards,

Marie