yade-users team mailing list archive

[chiara modenese <c.modenese@xxxxxxxxx>]

2010/4/15 Václav Šmilauer <eudoxos@xxxxxxxx>

> > 1) PWave time step (could you give me some references on this, please)
> >
> > 2) GSTS
> >
> > 3) A fixed constant value set by the user
> >
> > Now I am not sure that for the explicit scheme we are using and for
> > the case of non linear relationship between forces and displacements
> > (that is of my interest) we end up with the critical time step for the
> > system using one of the above. Has anyone tested this aspect before?
> > Do you have any references to suggest me that eventually touch this
> > issue (if any)?
> > I am not even sure why, if I use the GSTS, I should give the
> > tangential stiffnesses and not the secant values as input.
> > In summary, do you see any limit of applicability for the use of GSTS?
>
> GSTS computes the "real" critical time step (based on numerical
> stability of the integration scheme), whereas PWave only makes an
> estimate. Particle oscillation is governed by current stiffness (that's
> why you need tangent, i.e. current, stiffness).
>
> If you have variable stiffness, you should set GTST safetyFactor and
> interval to such values that you don't go over Δt even if running GSTS
> only sometime. Actually, if you go over Δt on one interaction for short
> time, you might not even notice in overall (explosion etc), as it will
> locally stabilize later...
>

Thank all for your answers.
I think this safetyFactor is probably what I will need if I do not want to
run GSTS at each time step. Now I do not know how to calculate this factor,
but I think it is a good idea if I want to save CPU time.
In fact, having non linearities, at each time step, not only the number of
contacts is changing but also stiffnesses differ all from each other since
their values actually depend upon the overlaps. To get right the non
linearity I suppose the time step will be very small. Maybe I do not need to
be worried about CPU time now, but I will find a way to calculate reasonably
such an interval to run GSTS.

Chiara



>
> (For reference, I send you working copy of my thesis separately... I
> know of no other ATM.)
>
> Cheers, Vaclav
>
>
> _______________________________________________
> Mailing list: https://launchpad.net/~yade-users<https://launchpad.net/%7Eyade-users>
> Post to     : yade-users@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~yade-users<https://launchpad.net/%7Eyade-users>
> More help   : https://help.launchpad.net/ListHelp
>