yade-users team mailing list archive

[Bruno Chareyre <question697174@xxxxxxxxxxxxxxxxxxxxx>]

Question #697174 on Yade changed:
https://answers.launchpad.net/yade/+question/697174

    Status: Open => Answered

Bruno Chareyre proposed the following answer:
(sorry for sending earlier answer to users list, here it is)

Hi,
The problem is you parameters are extreme and the servo control of boundaries doesn't like that:
Super-high stiffness with super low confinement. If you visualize interaction forces in the 3D view you can see that they keep blinking, that's because the walls struggle to maintain a constant stress and the force oscillates between zero and something.

The particular case also leads automatic timestep to not be correct (may also be linked to the viscous Mindlin law you are using).
Without changing the physical problem I could stabilize your script with:
timestepSafetyCoefficient=0.25
triax.max_vel=0.002

I would discourage generating a randomDensePack for a triaxial loading since, what it does is actually another triaxial loading. Then when you load it as an initial state you have a discontinuity in mechanical properties which make it unstable again. Especially in your super-stiff case. Just a waste of computation time.
Better do everything in same conditions.

I would also avoid using super low confinement-to-stiffness ratio (1e-6
in your case) if possible, since it makes the whole thing not very
efficient.

I'm surprized that randomDensePack seems to not reload previous packs
despite you using memoizeDb. Do you see the same behaviour?

I hope it helps

Bruno

-- 
You received this question notification because your team yade-users is
an answer contact for Yade.