problem using a 3D extruded mesh with periodic boundary condition

[lzhang]

Hello,

I have a test case involving the simulation of flow around a cylinder. For this test, I use 3D meshes extruded from a 2D one. For the first mesh with 10 cells in the z direction, when I run the simulation I got an error "*** Error in `./cs_solver': malloc(): memory corruption: 0x0000000001784ed0 ***". However the number of cells is not that large (about 270000 cells) to have a memery allocation problem. For the second mesh with 4 cells in the z direction, I got a different error "Cannot achieve to reorient the current joined face with face 45313 (selected face)."

The boundary condition for faces ("front", "back") whose normal is in Z direction is set to be periodic. Certainly, the periodic boundary conditions are not well set, since If I use Symmetry type for these two faces, I have no more previous errors.

Could you please tell me what is going wrong in the periodic BC setting, please? I'm using code_saturne 5.0.4, and in attached file you can find the .xml file, mesh files and the listing files.

Best regards,
Lei

[lzhang]

I add also .xml file and the mesh file for test case 2 (4 cells in Z direction).

[Yvan Fournier]

Hello,

There was a fix in version 5.0.6 in parallel mesh extrusion (which occured when some ranks had no boundary faces), so could you test if you still have this crash with version 5.0.6 or more (5.0.9 bug-fix was released today, so I recommend that one).

Regards,

Yvan

[lzhang]

Hello,

Maybe I was not well explaining my problem. I did not use mesh extrusion of code_saturne to build mesh, in fact the extruded mesh was already built with salomé. So I think it is a problem related to periodic boundary condition.

I will test the version 5.0.9 and let you know if the problem is solved.

Best regard,
Lei

[lzhang]

Hello,

I just tested the 5.0.9 version, and I encountered the same error.

Certainly, it is due to the BC boundary condition. Do you have any ideas to solve it, please?

Best regards,
Lei

[Yvan Fournier]

Hello,

I checked the first case, and it seems you defined the periodicty incorrectly: your mesh has a z range of [0, 0.07456], so your periodicity in z should be 0.07456, not 1 (it is not just the direction, but also the distance).

With the correct value, things seem OK. I assume the issue in the second case is due to similar causes (I'll let you re-check).

Best regards,

Yvan

[lzhang]

Hello,

Indeed it was a problem with the translation vector.

All works now!

Thanks a lot!

Best regards,
Lei