strange effects of Law2_Dem3DofGeom_FrictPhys_Basic in dense periodic packings

[Bruno Chareyre]

I made comparisons between Law2_Dem3DofGeom_FrictPhys_Basic and Law2_ScGeom_FrictPhys_Basic in periodic triaxial tests. Damping was very law (2%), 400 spheres with identical positions.

The result is attached. You'll see in gray the curve of unbalanced force for ScGeom, resp. pink for Dem3Dof.
Different stress curves is ok, we can get almost the same difference between identical simulations in multithread (here, everything is run in 1 thread).

The surprising thing is unbalanced force is exploding, even with very low strain rate, after enough deformation, even though the same unbalanced force can go as low as 1e-8 in the initial isotropic compaction. I could repeat that with different number of spheres and different dampings.
With ScGeom, unbalanced force stagnate at very low value (so close to x-axis that you just see a few picks), exactly as what we would have with rigid boundaries.

Carefull examination and comparisons with usual DEM results on spheres show that the evolution of stress and porosity is impacted by this growing instability (the inverse would be surprising).

I have no real explanation for this. The only thing comming to mind is ratchetting (guess why the doc. has been updated...) but I'd give it a 10% chance only.

[Bruno Chareyre (bruno-chareyre)]

It needs more comparisons to know when this happens exactly. Those results were obtained in r2255, with :

PeriTriaxController(dynCell=True)
NewtonIntegrator(damping=0.02, homotheticCellResize=True)

I didn't test without homothetic resize.