Triangle-grain force computation depends on numbering (Liggghts 1.5.x)

Submitted by Franzzzzzzzz on Tue, 02/26/2013 - 09:50

I've found a problem in Liggghts 1.5 and 1.5.3 when computing the force between a grain and a mesh triangle when the triangle size is of the order of the grain diameter. It appears that the force on the triangle depends on the way the triangles are numbered in the mesh. For example, when a grain slide on a square (2 triangles separated by the diagonal), the force on each triangle depends on which one was defined first in the imported stl (I can attached the input file if needed).
This problem seems to have been corrected in liggghts 2.2.4, and I don't know if the 1.5 branch is still maintained ; but since it is still available in download I think it could be important to at least underline this problem in the documentation.

Forums:

ckloss's picture

ckloss | Tue, 02/26/2013 - 23:22

Hi,

thanks for your bug report.
Yes - there has been quite some we-work in the 2.X branch to fix things like these. And to frankly answer your question: No, the 1.5 branch is no longer maintained. I don't think there are many people still using 1.5 and especially the mesh implementation has diverged too far so that fixes can't be ported back. To clarify this, I deleted 1.5 / 1.5.3 from the download section.

Cheers
Christoph

Franzzzzzzzz | Fri, 03/15/2013 - 17:23

Hi,
Thanks for the previous answer. So I switch to Liggghts 2.X (2.3 right now to be precise). As I said before, the precise problem that I described before disappeared, but I still have a problem with more complex geometry/more grains. To be more precise about that : I have a ycylinder centered on 0,0 ; grains above and below it without gravity and a top and bottom layer with springs on them to apply a mean pressure on the medium. The box is xperiodic and yperiodic. The grains are loaded from a restart file (read_data).
Since my problem is symetric in the x-y plane, I should be able to run the script twice with every atoms in the restart file with opposite z coordinate and get the same result (ie. running a xy-plane mirrored experiment). This is not at all what I see : the force on the cylinder quickly differs between the two runs, for mesh cylinder and also for primitive cylinder. I don't really see where is the problem...
(I would have put the input file but I don't see how to link a file here in the forum ...)

Cheers.