yade-dev team mailing list archive

[Václav S(milauer <eudoxos@xxxxxxxx>]

> Hi,
> I realized a new algorithm for facets. It simpler and faster then current one (it uses only the class Facet, without distinct classes Edge and Vertex), but it less exact (at least theoretically), and is not compatible with the current version (only triangular facets, facets se3 is at the inscribed circle centre). 	
> It less exact because contact with facet includes edges and vertices. But because the facets not linked, at each edge will double contacts (on the number of edge facets). The same for the vertices. 
> However, simulation of the rod penetration into a bed of sphere particles did not reveal fundamental differences in the total force on the rod with exact (current) and proposed algorithms (see ps for details). 
> So I think the accuracy of current algorithm is clearly redundant, and I propose to replace it with new one, rather than provide two algorithms. This would avoid confusion.
> That do all think about this? 
> Vaclav, what is your opinion (you seems use the facets)?
>   
Hi, I think that is a great idea. For the tests you run, there will be
not much difference since there is relatively few edges, but there can
be big influence of the doubled reaction on e.g. triangulated surface
with a lot of triangles to model the geometry details. I would suggest
adding some possiblit (flag?) to turn on/off reactions on each
particular edge and vertes - this way, there could an algorithm that
would enable only one of the coinciding edges etc. Does that make sense?

Nice work :-)

Vaclav
> ps. simulation details:
> Rectangular rod is translated at  0.075 m/s into a bed of  sphere particles. The mesh cell density on the rod is varied to analyse the  effects of a fine mesh and  course mesh on the calculated  forces. Total force on rod averaged around three simulations. I uploaded the relevant figs on wiki:
> http://images4.wikia.nocookie.net/yade/images//6/68/Rod-coarse.jpeg
> http://images1.wikia.nocookie.net/yade/images//6/69/Rod-fine.jpeg
> http://images3.wikia.nocookie.net/yade/images//0/00/Rod-total-force.png
>
> pps. proposed algorithm inspired by Kremmer, M. & Favier, J. F. A method for representing boundaries in discrete element modelling~--- part I: Geometry and contact detection Int. J. Numerical Methods in Eng., 2001, 51, 1407-1421
>