Mesh from salome - problem with convert2syrthes

[michal777]

Hi,
I've got problems with making .syr file for syrthes from salome. I've got syrthes 5, salome 9.11.0 (for ubuntu 22), linux mint 21.2 (fresh, installed on virtual machine).
I tried different things:
- compiling syrthes with references to med and hdf libraries provided with SALOME - result: "free(): invalid pointer"
- using libmedc from repository (it's loaded if I don't source syrthes.profile) - result: no error messages but the syr file is probably broken (syrthes complains about degenerated elements of mesh)
- using convert2syrthes from repository - result: output file empty, some error messages in french
- exporting different versions of med file from salome - no difference
- salome 9.9.0 instead of 9.11.0 - no difference
- using unv format instead of med - convert2syrthes doesn't accept it
- changing mesh in salome - triangular or tetrahedral - still the problems
- the attempts were mostly with the simplest possible mesh - a single 1x1x1 cube (but also with more complex meshes).
I'm not sure if salome uses it's own med and hdf (provided in BINARIES-UB22.04) or the installed from repository, I don't know how to check or change it.

Any ideas? Has anyone working med->syr conversion with some recent versions of the software?

[Yvan Fournier]

Hello,

Does you mesh contain tetrahedra and triangles only ? Otherwise, that might explain the empty file and error messages when using convert2syrthes from repository ?.

Also, are you using Syrthes for standalone computations of for conjugate heat flu with code_saturne ? In the latter case, you can use code_saturne's built-in "internal coupling" mode (simply defining some volume zones as solid) for most configurations, unless you need some features such as thin thermal resistances and such.

Best regards,

Yvan

[michal777]

I don't really know which features I need.
I saw a 3 2D disks tutorial and I thought that mayby something like that (but 3D) could be useful to do some simplified simulations of electronics (hot parts on PCBs, heatsinks, distribution of heat by copper planes, effects of convection). "Thin thermal resistances" may be needed but I don't understand what "thin" exactly means and what are limitations of "internal coupling" in comparison with syrthes.
If I understand correctly, code_saturne can simulate both conduction and convection without syrthes. Is there any documentations which explains how to do it?

[Yvan Fournier]

Hello,

Thin thermal resistances occur when there is an extra resistance at the interface of two materials, so you might have that in tour case. It is not handled yet in code_saturne, but should be relatively easy to implement and test.

Besides than, which could be added later, using the code_saturne GUI, you only need to setup the computation as for a fluid, then define which zones are solid (then assign specific physical properties to those zones). There is no detailed documentation, but not much else that is specific.

Otherwise, is most of you mesh fluid or solid ?
Also, you did not indicate what type of elements were contained in the file you tried to convert.

Regarding the 3D disks tutorial, I think it fails with Syrthes 5. I sent a patch to the Syrthes team some months ago relative to this, but have not seen a new release integrating this... 3D/3D coupling would work though, the issue is specific to the 2D mesh on the Syrthes side.

Best regards,

[michal777]

3 2D disks tutorial worked for me but I used the meshes provided for the tutorial.
After getting through the tutorial I tried to make a next step - simple 3D simulation, similar to the tutorial but the mesh issue stopped me.

To test convert2syrthes, I made very simple meshes - just a cube, meshed with netgen either as triangles or tetrahedrons. According to mesh information in salome they have only 1D edges, 2D cubes and 3D tetrahedrons (if present). Nothing worked for me so it looks like salome and/or syrthes is broken or incompatible at the moment.

[Yvan Fournier]

Hello,

Looking back at your original post,

- using libmedc from repository (it's loaded if I don't source syrthes.profile) - result: no error messages but the syr file is probably broken (syrthes complains about degenerated elements of mesh)

might be a good place to start. What does Syrthes say whan it complains ?

Regards,

Yvan

[michal777]

Attached
- the logs from syrthes when it gets mesh which seems broken
- input mesh files: cube_simple_tri.med (cube, only 12 triangles), cube_many_tri.med (cube with more dense mesh), Mesh_solid.med (more complex)
- results "convert2syrthes results med from salome" - results after source syrthes.profile (med, hdf libraries provided with salome)
- results "convert2syrthes results med from system" - results without source syrthes.profile (med, hdf libraries from system)
Looks like the .syr files have a lot of zeros at the end (seems wrong).

[Yvan Fournier]

Hello,

Converting the MED file from a local build of convert2syrthes on my machine (with an older Syrthes 4.3), I get a Solid.syr file with the same number of elements, the same coordinates, but completely different element to node connectivity than what you get with the system MED libraries.

Even with the system libraries, you should source "syrthes.profile" before running the conversion. Doing that, can you check the output of "ldd" on the convert2syrthes file, to make sure you are not pointing to another version of the libraries for some obscure reason, probably if you have libmedC from the repository but maybe not the matching headers, and the build mixes that with those of Salome or the Syrthes build ?

And also post your setup.ini file to see how that matches.

The different error messages you get point to some mixup in libraries, so ldd will help you see hat is really being used before running the converter.

Regards,

Yvan