yade-dev team mailing list archive

[François <francois.kneib@xxxxxxxxx>]

Hi Klaus,
And thanks for this work ! My answer below :

2016-06-10 14:43 GMT+02:00 Klaus Thoeni <klaus.thoeni@xxxxxxxxx>:

> Hi guys,
>
> just pulled some helper functions for cylinders and some related examples.
>
> I also created a script examples/cylinders/cylinderconnection-roots.py
> which
> is aimed to replace examples/chained-cylinders/chained-cylinder-roots.py. I
> couldn't get the box-gridConnection interactions to working, but I guess I
> am
> missing something.
>

​If you have a closer look at the cylinders behaviour, you will see that
the spheres are feeling the cylinders, but the cylinders are not influenced
by the spheres. This is due to an incomplete O.engines, the main ideas are :
- Law2_ScGeom_FrictPhys_CundallStrack() is not handling cylinder-sphere
contact, we need Law2_ScGridCoGeom_FrictPhys_CundallStrack()
- We always need Ig2_GridConnection_GridConnection_GridCoGridCoGeom() , or
an "avoidSelfInteractionMask" correctly set if we don't want
cylinder-cylinder interaction. (the first solution will avoid contact
between two adjacents cylinders in the same chain and the second one will
avoid all connection-connection interactions).
I commited the corrected example script.
Cylinder-Box interactions are not supported but are not difficult to
implement. The reason in a few words :
- GridNodes have no bounding-box, we can't expect them to interact
- GridConnections interactions need specific law2, because they dispatch
the force on the nodes. For this reason, and similarly to the first issue
above, the wall will feel the cylinders, but the cylinders will not feel
the walls.
A better implementation would be a kind of cylinder-node interaction that
dispatches the force in all cases. This way the creation of
cylinder-specific laws would not be necessary, but I'm afraid it would
affect the performence a little.

In addition, have a look at examples/cylinders/cylinder-cylinder.py. Is
> there
> a way of linking the rotation of the GridNodes to the
> GridConnection/Cylinder?
> I guess increasing the bending stiffness would help but how can we make it
> independent of this? Any suggestions?
>

​This is not possible, as this bending rotation is the way an assembly of
connected cylinders are bending. If we block this rotation, we block the
bending of the gridNodes/gridConnection. To better understand, always keep
in mind that the grid internal behaviour is basically made of remote
cohesion interaction, and here the rolling moment between 2 spheres is the
way we make the bending of a beam. This behaviour is strange because you
use mono-connection grids, and you look at the extremity . In real
simulations, thoose beams should be spacially discretized unless you want
to model a tic-tac box ... and a tic-tac is much stiffer than the example
you made I guess :-D


> Cheers
> Klaus
>

​Again, thanks for your work.
​Hope I was clear, don't hesitate to ask me more if you need. Now I come
back to my PhD reality :-P​

Cheers,
François


> On Tue, 17 May 2016 09:31:13 AM François wrote:
> > 2016-05-11 15:11 GMT+02:00 Bruno Chareyre <
> bruno.chareyre@xxxxxxxxxxxxxxx>:
> > > On 05/11/2016 01:52 PM, Klaus Thoeni wrote:
> > >> I think sooner or later we should get rid of the current
> ChainedCylinder
> > >> implementation.
> > >
> > > Agreed
> > >
> > >> But you are right, the algorithms are slightly different,
> > >> although it might just the physics (?). I haven't compared the code
> yet
> > >> but I
> > >> added two examples in examples/grids which are the same as in
> > >> examples/chained-cylinders and the results are slightly different.
> > >
> > > Mmmhh.. I'll have to have a look.
> >
> > Sliding contact ​tracking algorithm is slightly different so the results
> > are a little different too.
> >
> > >   Any idea how
> > >>
> > >> many people are using ChainedCylinder?
> > >
> > > Not many AFAIK
> >
> > I converted everyone to grid here at Irstea.​
> >
> > > Bruno
> > >
> > >
> > >
> > >
> > > _______________________________________________
> > > 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
>
>
> _______________________________________________
> 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
>