yade-users team mailing list archive

[Bruno Chareyre <question706553@xxxxxxxxxxxxxxxxxxxxx>]

Question #706553 on Yade changed:
https://answers.launchpad.net/yade/+question/706553

Bruno Chareyre posted a new comment:
Hi,
I think strain damping could give a null volumetric rate in this case. It's consistent with the fact that even with 0.99 ev is still (somehow) small.

The algorithm implies an exponential convergence to the goal rates, approximately like this:  
      rate += (goal-rate)*(1-damping) (a discrete form of d(rate)/dt = - a*(goal-rate))    [1].
If I'm not wrong, it leads to proportional increases of the rates in the different directions, and so the trace of the strain rate tensor should be always zero.

In this reasoning I'm assuming that all strains start from zero, though. If you start the constant volume stage from a state where the velocities are not zero (which is the case if you impose stresses for instance), then there can be small offsets of the different components during the convergence. 
If that is the cause of the small volume changes, there could be multiple solutions, e.g.
- assign goals=0 and run a few iterations, then shear
- assign velocities=0 directly (wall.state.vel=(0,0,0)), then shear

Another option is to simply set set strainDamping=0, and implement your
own ramp in the script, possibly reproducing [1] or with a different
equation, changing the goals incrementally (this is to reduce elastic
waves mainly).

Bruno

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