yade-users team mailing list archive

[chiara modenese <c.modenese@xxxxxxxxx>]

Hi Bruno!

On 4 October 2010 14:47, Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx> wrote:

> Hi again,
> Vaclav told me this answer was not fully satisfying. If the question was
> "why ratcheting has been implemented in yade in the first place", I'm not
> the good person to answer. My best guess is it was inherited from SDEC,
> which was itself more or less reproducing PFC3D.
>
Thanks for replying again. Yes, my question was: is it right to derive the
branch vector from the position vector of the contact point rather than
getting it from the radius*normal? I am not interested in the history of the
code, I am more concerned with the correctness of the code in principle. I
think that, at least for the Hertz-Mindlin case, it is better not to choose
the contact point because, in fact, we deal with elastic deformations at
contact, where there is no overlap.



> I don't think it can give big differences in the most usual situations, and
> I suspect explosions in periodic BCs have another explanation.
>
Concerning the periodic case:
-- I remember that you tested the code and said that you were comparing the
results with the ones obtained from rigid boundaries conditions. Have you
done those tests avoiding ratcheting?
-- I looked again at the way we compute the relative velocity at contact. I
do not see anything wrong in the way we apply the shift due to the change in
the velocity gradient. The explosion of the sample happens only for the case
with ratcheting, though. It is strange, I agree, but we should find an
explanation. Only one doubt for the moment: we update relative velocity at
contact and then we also update the velocities in the NewtonIntegrator
class. Is this correct? I know it is an old question but I still have the
feeling that we are updating the velocities twice (I recall that Cundall is
only updating particle positions and then relative velocities).

Thanks in advance.

Cheers, Chiara

PS @Vaclav: True :-)

>
> Cheers.
>
> Bruno
>
>
> On 30/09/10 16:35, Bruno Chareyre wrote:
>
>> Hi Chiara,
>>
>>> allow me a question about the way we compute the shear part. Following
>>> the Cundall's model, no problem of ratcheting would arise since he defines
>>> the branch vectors as radius*normal.
>>>
>>
>> I recently checked PFC manual, and it defines shear with OC vector,
>> resulting in ratcheting. Maybe you'll find radius*normal in other Cundall's
>> papers, not sure, but at least PFC3D gives ratcheting.
>>
>>  Now in Yade if we avoid granular ratcheting we follow exactly the same
>>> way. So, apart from ratcheting, why should the current length of the
>>> interaction be preferred? Why this was introduced in the code?
>>>
>>
>> I think it is often introduced by DEM programmers because it is somehow
>> intuitive, once you admit that particles can overlap each other, with a
>> "contact point" somewhere in the overlap. Current length is also used in
>> torque definition, whatever the definition of shear displacement.
>> Personally, I try to never use the words "overlap" or "penetration", and
>> refer to relative displacement of centers instead. The "overlap" concept is
>> not needed to derive the equations and it misleads people (another example
>> is the computation of void ratio : people sometimes tend to remove the
>> "overlapping" volume, which not suitable in most cases IMHO).
>>
>> Bruno
>>
>>
>
> --
> _______________
> Bruno Chareyre
> Associate Professor
> ENSE³ - Grenoble INP
> Lab. 3SR
> BP 53 - 38041, Grenoble cedex 9 - France
> Tél : +33 4 56 52 86 21
> Fax : +33 4 76 82 70 43
> ________________
>
>
> _______________________________________________
> 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
>