yade-users team mailing list archive

[chiara modenese <c.modenese@xxxxxxxxx>]

>
>
> If bodies from different periods are in contact, you need to translate at
> least one of them to its ghost position in order to define the local
> geometry, or (b->pos - contactPoint) can be anything, even larger than the
> period itself. Actually, in periodic ScGeom we have no position and no
> contact point (at least we don't use them).

We do not wrap the positions but we wrap the contact point, although we do
not use once we avoid granular ratcheting.

>
>  Related to HomotheticResize.  May I know the difference between velocity
>> and displacement update for HomotheticResize?  Cundall and Thornton use
>> displacement update.
>>
> The way positions were updated was not correct, it was just a quick hack.
> After looking at the different facets of the problem for some time, I
> concluded that updating velocity was more elegant and the implementation was
> way simpler (it fits in just one line in Newton).
> If you update positions, it raises questions on what is the true velocity
> (and you definitely need correct velocities for viscous laws or incremental
> elastic-plastic laws). OTOH if you update velocities, it is reflected in
> positions automatically. It should be equivalent with Cundall.
>

Hello Bruno,
I am a little bit confused about the periodic. In NewtownIntegrator we
update the velocities and _then_ we integrate the positions. This is exactly
the way of Cundall (I am checking the equations right now) except that he
chooses to update positions and not velocities (defining a strain rate
instead of a gradient of velocity which we said is fine but not in the
general case). I understand that the reason why Cundall updates the relative
velocities at contact (btw, according to his paper this is not meant to be
only for the shear direction) before going for the computation of contact
forces is that he was not updating the velocities at the level of the
integration. So my guess is: why do we need to update the relative
velocities if that has already been done in NewtonIntegrator? That would
have made sense if updating positions and not velocities. Bruno, you said
there was a problem using periodic ScGeom before the last changes. Where you
testing with homothetic=1? Because I think that case you would have needed
to update velocities (which is not a problem in Dem3dofGeom).

Thanks in advance for your explanation,
Chiara


>
> Bruno
>
>
>
>
> _______________________________________________
> 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
>