yade-dev team mailing list archive

[Václav Šmilauer <eudoxos@xxxxxxxx>]

> Haha, Vaclav're quick.  I've installed 1853, and u've uploaded 1854.
> When I run simple-scene.py, it couldn't run, and the Player doesn't
> seem to come out even if i press F9-F12.  Only the generator comes
> out.
What did it say for simple-scene.py? It works just fine here.

> Q2. Another novice question is after merging with bazaar, should I
> commit to my local system by typing "bzr -commit -m 'abcd' "?

Contrary to what Sega recommends, I work just with the main trunk and
update whenever someone else commits; that minimizes potential
conflicts.

Whether "bzr commit" is local or not depends if you did a checkout from
lp:yade  or if you branched. In the latter case, you need to call "bzr
push" to get your local commits to launchpad. You can also use bzr
gcommit (gtk interface to commit) if you have the bzr-gtk package
installed.

> Q3. Have the reported bugs on clumpId, clump::updateProperties in this
> website "http://beta.arcig.cz/~eudoxos/yade/doxygen/"; been solved?

Which ones? (you pasted the whole documentation URL, you have to get the
inner frame URL) If you men the eigendecomposition, it is worked around
later in the code by detecting NaNs in the matrix. So, yes, it generally
works. I wrote those bug notices almost 3 years ago and there were no
complaints till now (there were not may people using clumps either, but
that seems to be changing now).

> Q5. Is there a C++/python example code on how do I retrieve data (with
> graph plotting) from a Triaxial Test?  E.g. strain, principal
> stresses, deviatoric stress, and orientation/rotation distribution.

There is scripts/simple-scene-plot.py (I just fixed it) that uses simple
interface to track variables. For triaxial, it would be feasible in
python as well, but for now you can just use TriaxialStateRecorder,
which creates text file suitable for e.g. gnuplot.

> Q6. I looked at LSMgengeofile to generate particles. Is the
> distribution between rmin & rmax completely random?  Why is the
> particle generation technique used in Triaxial Test not recommended,
> as was mentioned by Bruno?

Packing generation is complex subject and I want to devote whole chapter
on it in the docs I am writing.

There are static (geometrical) procedures to create packing: trivial for
regular packings (pack.regularHexa and pack.regulaxOrtho), but
complicated if you need something random (isotropic, with at least
approximate size distribution etc): this is where you might find LSM
GenGeo useful (you have to install
http://www.launchpad.net/esys-particle, it is not part of yade). Then
there is Jean-Francois Jerrier's algorithm, which is not integrated in
yade fully yet (he promised to do that around Christmas IIRC) and uses
random tetrahedral mesh for start.

Then there are dynamic procedures, which basically run DEM simulation
with special boundary conditions (such as triaxial compression and
relaxation). They might take longer (sometimes), but you can prescribe
more properties (such as size distribution). The disadvantage is that
the packing is not "stress-free" (as they call it, misleadingly a little
bit), i.e. there are slight overlaps of spheres; OTOH you can define the
stress-free positions to be those intial ones, with slight overlaps, and
that works around the problem very nicely. For those, see
http://beta.arcig.cz/~eudoxos/yade/epydoc/yade.pack-module.html#randomDensePack

Have a look at
http://beta.arcig.cz/~eudoxos/yade/epydoc/yade.pack-module.html in
general, notably examples mentioned at the top of that page. That might
save you lots of time.

Cheers, Vaclav

> Thanks Vaclav for writing the API and tutorials...
(You're welcome :-) This is suspended for a few days now, as I have to
do some real research work...)