Postprocessing fails

[CsatiZoltan]

Hi,

I modified the 1_uniaxial example for my grain structure. I did not tessellate the geometry by Neper, but obtained the necessary .msh file in another way. FEPX completes the simulation but Neper fails with segfault during the visualization.

========================    N   e   p   e   r    =======================
Info   : A software package for polycrystal generation and meshing.
Info   : Version 4.1.2
Info   : Built with: gsl|muparser|opengjk|openmp|nlopt|libscotch (full)
Info   : Running on 8 threads.
Info   : <https://neper.info>
Info   : Copyright (C) 2003-2020, and GNU GPL'd, by Romain Quey.
Info   : No initialization file found (`/home/zoltan/.neperrc').
Info   : ---------------------------------------------------------------
Info   : MODULE  -S loaded with arguments:
Info   : [ini file] (none)
Info   : [com line] .
Info   : ---------------------------------------------------------------
Info   : Reading input data...
Info   :   - Reading arguments...
Info   : Writing simulation directory from FEPX raw result directory...
Info   :     [o] Writing directory `/home/zoltan/Programs/FEPX/examples/my_example.sim'...
Info   :   - Parsing FEPX results...
Info   :     > Node number      : 2790
Info   :     > Element number   : 1327
Info   :     > Elset number     : 0
Info   :     > Partition number : 1
Info   :     > Step number      : 4
Info   :     > Node results     : coo
Info   :     > Element results  : strain-el stress
Info   :     > Elset results    :
Info   :   - Writing inputs...
Info   :     [o] Writing directory `/home/zoltan/Programs/FEPX/examples/my_example.sim/inputs'...
Info   :       . simulation.msh...
Info   :       . simulation.config...
Info   :       . *.sh...
Info   :     [o] Wrote directory `/home/zoltan/Programs/FEPX/examples/my_example.sim/inputs'.
Info   :     [i] Parsing file `/home/zoltan/Programs/FEPX/examples/my_example.sim/inputs/simulation.msh'...
Aborted (core dumped)

It seems that Neper doesn't like my mesh file. What could be the problem? I attached the necessary files (renaming them to .txt so that GitHub accepts them).

simulation.msh.txt
simulation.config.txt

Thanks in advance,
Zoltan

[mkasemer]

Hi Zoltan,

Thank you for your interest in FEPX. Since the mesh was not made with Neper, it may be difficult to find the precise issue - that is, something we could recreate (using Neper commands) and check to see why the code was failing. However, it sounds like the simulation completed successfully, indicating that the mesh was proper. Can you please confirm by copying the terminal output from the simulation?

Otherwise, this is a problem for the Neper Discussion Forum, as this is an issue with Neper, not FEPX.

[mkasemer]

Concerning BCs: This is true. I will try to do the same on my end for sake of comparison.

Concerning loads: you are applying strain steps, so domain size should not matter. The loads should be quite large, as your domain is rather large.

[CsatiZoltan]

@mkasemer
I managed to strip down the example mesh, where only the volumetric elements are kept (see attached). The simulation runs successfully (I even tried it with loading in the Y direction to have an additional test). The reason I wanted to try this is to make sure that

• no explicitly given BC is necessary
• surface face information ($Fasets) is not vital in this case
• it resembles my .msh file (also attached).

Now my simulation uses a domain similar to the unit cube in your example, the residuals are of the same magnitude (1E-2).
The first step of the visualization script (neper -S .) runs successfully, but the visualization step still fails when Neper parses the mesh. It seems that this is a mesh-related issue, so I opened an issue on the Neper side.

simulation.msh.txt
3D_testFEPX.msh.txt

[mkasemer]

Thank you for the update, Zoltan. Are you confident in the output results? Are you getting believable values for stress/strain?

[CsatiZoltan]

The magnitude of the strain and stress components are of the same order for my case as for the built-in example 1_uniaxial. Although I have only 2 grains, not 100, the material parameters are the same (I copied the config file) and I have no grains with sharp features, so yes, I would say that the numbers are believable. Of course, I would like the visualization to work in order to check the displacements.

[CsatiZoltan]

@mkasemer It turned out that it was a Neper issue with the mesh topology, and has been resolved. I do not see where I could set the state of this thread Answered, but probably you can do it.