Clumps are getting a very high angVel in one iteration

Is there something special hapening to state-ori before step 154675 (manipulation in the script, some specialized clump functions, etc?), because orientation quaternion seems very badly normalized.

A rotation matrix should be skew-symmetric, here you have:
A: 0.934319 -0.193747 0.299181
 0.188102 0.980983 0.0478483
-0.302762 0.0115711 0.952996

I'm not sure it is related to your problem, but it is very strange. Orientation is normalized at the end of leapfrogAsphericalRotate, so it should be ok at the next step. But again at step 154676 it does not give a good rotation matrix.
Something happening at each step, before Newton?!

2014-05-19 11:56 GMT+02:00 Bruno Chareyre
<email address hidden>:
> Bruno Chareyre proposed the following answer:
> Is there something special hapening to state-ori before step 154675
> (manipulation in the script, some specialized clump functions, etc?),

No, it is just a free fall of clump and it reaches the floor at the step 154675.
Some more info for previous steps:

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
b->id: 209; Iter:154673
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
A: 0.934319 -0.193747 0.299181
 0.188102 0.980983 0.0478483
-0.302762 0.0115711 0.952996
....
....
state->ori: 0.9834 0.00922239 -0.153026 -0.0970739
state->angVel: 0 0 0

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
b->id: 209; Iter:154674
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
A: 0.934319 -0.193747 0.299181
 0.188102 0.980983 0.0478483
-0.302762 0.0115711 0.952996
M: 0.934319 -0.193747 0.299181
 0.188102 0.980983 0.0478483
-0.302762 0.0115711 0.952996
....
....
state->ori: 0.9834 0.00922239 -0.153026 -0.0970739
state->angVel: 0 0 0

> because orientation quaternion seems very badly normalized.
>
> A rotation matrix should be skew-symmetric, here you have:
> A: 0.934319 -0.193747 0.299181
> 0.188102 0.980983 0.0478483
> -0.302762 0.0115711 0.952996
>
> I'm not sure it is related to your problem, but it is very strange. Orientation is normalized at the end of leapfrogAsphericalRotate, so it should be ok at the next step. But again at step 154676 it does not give a good rotation matrix.
> Something happening at each step, before Newton?!

Hmm, no there is nothing especial. I did not change the code either, as I never
worked with clumps. Interesting is that this clump is not the first one in the
simulation.

Anyway thanks a lot for info! That is something already.

Anton

Clumps are generating in this script continuously. And before explosion
happens, there is already in the scene about 15 clumps and there are no
problems with those.

The clump #16 reached the floor and "jumps away" from simulation. Very
strange...

Anton

2014-05-19 13:26 GMT+02:00 Anton Gladky <email address hidden>:
> Hmm, no there is nothing especial. I did not change the code either, as I never
> worked with clumps. Interesting is that this clump is not the first one in the
> simulation.

Bad mistake, sorry: rotation matrix are not anways skew-symmetric.
The quaternion you show is normalized, as it should. Forget #1...

Hi Anton,

do other clumps also interact with the floor? is it possible to check how
much "excentric" is the contact force, i.e. what is the contribution of
shear force? or maybe to check ratio of resulting torque and inertia..

Could you also try to reduce time step (just before the problem) even more?
Although you wrote that reducing time step does not solve the problem.. it
seems to me like stability problems due to time step..

cheers
Jan

2014-05-19 13:46 GMT+02:00 Anton Gladky <
<email address hidden>>:

> Question #248869 on Yade changed:
> https://answers.launchpad.net/yade/+question/248869
>
> Anton Gladky gave more information on the question:
> Clumps are generating in this script continuously. And before explosion
> happens, there is already in the scene about 15 clumps and there are no
> problems with those.
>
> The clump #16 reached the floor and "jumps away" from simulation. Very
> strange...
>
> Anton
>
>
> 2014-05-19 13:26 GMT+02:00 Anton Gladky <
> <email address hidden>>:
> > Hmm, no there is nothing especial. I did not change the code either, as
> I never
> > worked with clumps. Interesting is that this clump is not the first one
> in the
> > simulation.
>
> --
> You received this question notification because you are a member of
> yade-users, which is an answer contact for Yade.
>
> _______________________________________________
> Mailing list: https://launchpad.net/~yade-users
> Post to : <email address hidden>
> Unsubscribe : https://launchpad.net/~yade-users
> More help : https://help.launchpad.net/ListHelp
>

Hi Jan,

2014-05-19 14:02 GMT+02:00 Jan Stránský <email address hidden>:
> do other clumps also interact with the floor? is it possible to check how
> much "excentric" is the contact force, i.e. what is the contribution of
> shear force? or maybe to check ratio of resulting torque and inertia..

All clumps are "landing" in the same place on the floor. And only 16th clump
after reaching the floor flies away as well as all following ones
(e.g. 17th, 18th..). Interaction with other clumps as well as initial
overlapping is excluded. I tried \mu=0 (friction),
setting o.dt=0.005*PWaveTimeStep. The result is the same.

I was trying to push the same clump with a really high start velocity
and it works just normally. No NaNs.

It is getting funny :)

Thanks all for the help

Anton

There was a bug. Fixed here [1].

[1] https://github.com/yade/trunk/commit/4ea76ad6e47ac5074a389ad61712a0840e8560a5

Thanks all for discussion

Anton