yade-dev team mailing list archive

[Jan Stránský <_honzik@xxxxxxxxxx>]

Hi,

>first of all I suggest that you change your e-mail name to read Jan (not
>Honza) since people not familiar with the Czech language must be
>confused about your double identity.
changed :-)

>See Shop::normalShearStressTensor. To make it independent on IGeom type
>and working for both periodic and aperiodic boundary, you must compute
>the distance from particle's position where the second particle is moved
>by the "shift2", which is something like
>scene->cell->Hsize*I->cellDist.cast<Real>(). (cellDist is a Vector3i,
>that is why it must be cast to Real (Vector3r) for the matrix-vector
>product to be syntactically correct).
Ok, the next time I will read the comments more carefully :-)

>We were actually just thinking, since there are now 4 implementations of
>computing stress tensor (one by Bruno in PeriTriaxController, another
>one by you in Peri3dController, your last one in
>Shop::stressTensorOfPeriodicCell and the last one by us (following
>Thornton) in Shop::normalShearStressTensor), it would be nice to show
>analytically that they are identical by doing some index gymnastics.
>Jan, you are very good for this kind of things, could you try to do
>that?
I know that the actual definition in peri3dController and in Shop::stressTensorOfPeriodicCell is the same, only written in different ways (index and compact notation :-) I can try to compare them all, but probably not today.. but in a few days yes, I will inform you :-)

>If they are the same (they should be), we could make one single
>implementation; ideally, it would be able to provide both normal/shear
>and summary tensor.
good idea

>A choice left for the user is whether one takes the current contact
>length (particle distance) or the "reference" contact length (r1+r2 for
>ScGeom, refLength for Dem3DofGeom) as the particle distance. I don't
>know if some has ever compared results from both; and it has to do with
>the unfamous granular ratcheting as well.
At least in Dem3DofGeom, the choice is on type of simulation (small x large strains). refLength is probably faster, but not usable in large strains..

have a nice day

Jan