Convergence Problem to Simulate Natural Convection (Buoyancy Driven Flow) Simulation

[Alexandre Douce]

I converted the mesh to the unv format.  You just have to change the mesh name to cavity_hex_2d.unv in the GUI. :)

[Mita Sarkar]

Hi Alexandre,

Thanks a lot for your continuous help. This time also Saturne gave me error in preprocessing stage (to read .unv file attached by you). When I check the mesh it gives me error. The error message is attached here. Please let me know why it is giving this error.

I am using gmsh to create geometry and mesh and I am not able to create pure hex mesh. If you know how to make pure hex mesh using gmsh, please let me know. I was wondering why the solution is not converging for tet mesh but converging for hex mesh (you mentioned it is converging for your hex mesh), is that a limitation of Saturne?
 
Thanks and Regards

Mita

[Yvan Fournier]

Hello,

I just tested Alexandre's mesh using our build of Code_Saturne 2.0-rc1, and I do not reproduce your crash (running on an intel Xeon E5504, Linux2.6.31 kernel, glibc 2.3.6, gcc 4.2.1 used for the Preprocessor compilation).

Could you check if the file was correpted on download ?

"md5sum cavity_hex_2d.unv" should return a1d9221e3200698729ec80d9ed90d8eb

Otherwise, you may use addr2line -e cs_preprocess <adress> (with adress values taken from hte call stack in the error message you posted) to help where the crash occurs.

Best regards,
 
  Yvan Fournier

[David Monfort]

Hello,

Have you downloaded the mesh from Windows? In this case, you could have some "spurious" characters at the end of each line. We should have handled this issue for long, but well... maybe the bug reappeared ;)

If this is the case, try to run dos2unix <file> in order to get rid of these Windows characters (I don't remember how to install this tool...).

David

[Mita Sarkar]

Hi David,

I did not download the file from Windows. I downloaded the file on my Linux machine, extracted and used it and it is giving the error that I posted. If you have any suggestions for this, please let me know.

Thanks and Regards

Mita

[Alexandre Douce]

Please, could you return the results of the command md5sum cavity_hex_2d.unv?

Do you try addr2line?

Could you with GMSH build a real 2D mesh of triangles and then extrude it with a single element in the thickness in order to have a mesh made of prism?

[Mita Sarkar]

Hi Alexandre,

There might be some problem while I am trying the command md5sum cavity_hex_2d.unv. Is there any particular location from where I should run this command or I can run it from my study directory? When I run this command, it says "md5sum: cavity_hex_2d.unv: No such file or directory". So, I do not have any results of this command.

I tried 'addr2line' also and it says "bash: syntax error near unexpected token `newline'".

No, I did not create any 2D mesh and extrude it. Instead I created 6 surfaces, one volume and set the target element size. In that way I made sure that in 'Z' direction, there is only one element. If I use extrude surface, I faced difficulty to define my boundaries. Anyway I will try it again by extruding the 2D surface as you suggested. If you have any other suggestions, please let me know.

Thanks

Mita  

[Yvan Fournier]

Hello,

to run md5sum on your mesh file (to check for damage to the file), just run it in the directory where the file resides, passing the file name as an argument.

Regards,

  Yvan

[Mita Sarkar]

Hi Yvan,
 
I ran md5sum on my mesh file and it returns  a1d9221e3200698729ec80d9ed90d8eb  cavity_hex_2d.unv. What does it mean? Is that mesh file corrupted?

Thanks and Regards

Mita

[Yvan Fournier]

Hello,

You obtain the same result as we do ith md5sum, so your file does not seem corrupted, which means the issue is with the Code_Saturne preprocessor.

As we are unable to reproduce this issue on our side, it may be due to a compiler or library issue. What hardware and software (OS version, compiler versions, ...) are you running on ?

Regards,

  Yvan