← Back to team overview

yade-dev team mailing list archive

[Bug 1301443] Re: Spheres/Box contact : spheres ori suddenly set to NaN.

 

I have an idea... maybe.
Do you see a numerical instability with angVel changing sign at each step before it becomes NaN?

-- 
You received this bug notification because you are a member of Yade
developers, which is subscribed to Yade.
https://bugs.launchpad.net/bugs/1301443

Title:
  Spheres/Box contact : spheres ori suddenly set to NaN.

Status in Yet Another Dynamic Engine:
  New

Bug description:
  Hi,

  Please see and launch the attached script.
  Some spheres are falling into a box because of gravity. The simulation starts well, but when all particles are almost steady, wait for a few seconds and you will see that some of them who are in contact with the bottom wall will disapear. I spent some time to investigate and that's my preliminary conclusions :
  - it is fully reproducible
  - particles don't really disappear : their orientation is suddenly turned to Quaternion(Nan,Nan,Nan,2*pi). It's not linked with a high torque on them. All other attributes remains okay, and the other particles still can interact with it. If you manually set the orientation to a correct value after it disappears, it will appear again at the right place.
  - you can reproduce this with one sphere (uncomment the last line of the script and comment the previous paragraph) : let it fall then apply to it a little angVel : "O.bodies[-1].state.angVel=(0.,0.,0.001)". It will disappear after a few timesteps.
  - it's independent on the timestep value.
  - it's dependant on the interpenetration of the contact : if you increase the gravity or decrease the young modulus, you will be able to avoid this behaviour.

  Does somebody have an idea ? I think it's a rotation calculation
  issue, in a very specific geometric case.

To manage notifications about this bug go to:
https://bugs.launchpad.net/yade/+bug/1301443/+subscriptions


References