yade-dev team mailing list archive

[Chiara Modenese <c.modenese@xxxxxxxxx>]

Hi Bruno and Vincent,

thanks for your answers.

I have PFC manual on my desk (though I never directly used that code) and
they do define also contact viscous damping. In order to modify the time
step to account for damping effects, they re-define the tangential and
normal stiffness such that their values are increased by some factors.
Hence, the critical time step for viscous contacts should result be smaller
than the one for non-viscous contacts (do you agree?). I could not find out
where the PFC formulas come from, that is why I asked to you about that so
to have an alternative.

I appreciate your help (Sergei, perhaps you have also been using viscous
contacts, any ideas about the critical time step?).

Chiara





On 23 February 2011 16:43, Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>wrote:

>  X <#12e5365814c1bb85_>LatexIt! run report...
>
> *** Found expression $$a\ddot{X}+b\dot{X}+X=0$$
>
>  Hi,
>
> If this can't be found in numerical methods literature (but it should in
> fact), I think I could try and derive the stability condition for 2nd order
> explicit integration of a 1D oscillator ruled by:
>
> [image: $$a\ddot{X}+b\dot{X}+X=0$$]
>
> PFC will not help since they don't have viscous contacts (else, if they
> tell something about it, why not use it?).
> Actually, I have a colleague doing continuum modelling of solid-fluid
> couplings who would be glad also to know the answer, since he has very
> similar equations.
> When this first aspect is solved, GSTimeStepper is certainly the good place
> to sum viscosities of all contacts on one body, i.e. determine the "b"
> factor above. For now, it only defines the "a".
>
> Bruno
>
> p.s. Vincent, you defined viscosity in order to keep critical time-step
> below a given value IIRC. What was the reasoning behind?
>
>
>
> On 22/02/11 19:35, Vincent Richefeu wrote:
>
> Hi,
>
> Choosing an appropriate time step requires a certain expertise. Maybe a good reference could be the book of Allen and Tildesley (?) Keep in mind that the smallest time step depends also on the integration scheme. For instance, bigger time step can be used with velocity Verlet,Runge-Kutta or Beemans schemes when compared with leap-frog used in yade (I'm not sure about what I say).
>
> For a given integration scheme, let say that the critical time step determined for non-viscous contact will be smaller that the effectively minimum time step... Not sure it helps!
>
> Vincent
>
> Le 22 févr. 2011 à 18:12, Chiara Modenese a écrit :
>
>
>  Hi guys,
>
> I think we should introduce a limit to the time step for contact models using viscous damping. Would you have a good reference about that apart from PFC manual? Also would the GlobalStiffnessTimeStepper engine be a good place for such an introduction?
>
> Chiara
>
>
>
> _______________________________________________
> Mailing list: https://launchpad.net/~yade-dev
> Post to     : yade-dev@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~yade-dev
> More help   : https://help.launchpad.net/ListHelp
>
>