yade-dev team mailing list archive

[Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>]

Hello

I'm thinking to how MicroMacroAnalyzer (MMA) could be modified in order 
to be used in a more versatile way (partly because Vaçlav told me he 
could use that - still true?).

Currently, if you put MicroMacro engine in your simulation, 
MMA::action() will (in short) :
- scan bodies and build a regular(*) triangulation of the spheres each N 
iterations;
- compute the average strain in each tetrahedron and assign it to grains 
(so that the "strain" of one grain is a sum over adjacent tetrahedra), 
based on the dispacements of particules on the interval [n,n+N].
- save the state of the sample in a text file (with a specific 
convention), so that strains can be recomputed later (out of yade) with 
different interval sizes (like [n,n+2N], [n,n+3N], etc.)
- write local strains, porosity, fabric tensor, and other 
micromechanical quantities in txt files too.

Problems :
- This engine is doing always the same thing, you can't change what it 
does if you don't modify the code and recompile. It is a pitty, since 
many geometrical functions available in lib/Triangulation could be 
usefull as well (e.g. computing the dual Voronoi graph).
- If you want to access to quantities like porosity at the scale of 
tetrahedra, you have no way to do that in yade, because you need to 
access the CGAL structure using the concepts of cell_iterators, 
edge_iterators,... which are not defined in yade.

So, I'm wondering how to improve that, and I hesitate between 2 approaches :

a) enrich the interface of TesselationWrapper or MicroMacro, so that 
instead of just "MM::action()", many members could be accessed via 
python, and different things can be computed on demand, and returned in 
a standard format.  e.g. for the porosity in a cell of the 
triangulation, there could something like :

pair<int [4], Real porosity> MMAnalyzer::porosityInNextCell (void);// 
usage : while (porosity!=-1) porosity = porosityInNextCell().second;


b) expose and make full usage of all typedefs defined in def_types.h. In 
that case, no need for interface any more, the full lists of edges, 
cells, or facets can be visited, but the user need to learn how to use 
CGAL and/or my triangulation classes before doing anything. Example :

for (finite_cell_iterator cell=MMA->triangulation->finite_cells_begin; 
cell != MMA->triangulation->finite_cells_end; cell++)
{
     porosity = MMA::Porosity(cell);
     if (cell->vertex(2)->id == x) do something;//etc.
}

I created a draft wiki page about this : 
https://yade-dem.org/index.php/Triangulation

Bruno


(*) Regular is for weighted points (center + radius), while the usual 
Delaunay is for points. It can make a big difference (and Delaunay can 
give stupid result) when radii are not all the same.

--

_______________
Chareyre Bruno
Maître de Conférences

Grenoble INP
Laboratoire 3SR - bureau E145
BP 53 - 38041, Grenoble cedex 9 - France
Tél : 33 4 56 52 86 21
Fax : 33 4 76 82 70 43
________________