yade-users team mailing list archive

[Robert Caulk <question660585@xxxxxxxxxxxxxxxxxxxxx>]

Question #660585 on Yade changed:
https://answers.launchpad.net/yade/+question/660585

Robert Caulk posted a new comment:
Thank you Luc, Bruno, and Timos for responding to my question.

Hmmmm, I see now. OK so we are using a Delaunay triangulation to build
the facet orientations, right? Extrapolating Timo's comment, small
deformations should not change this triangulation, and we should be safe
to update positions. As soon as a large deformation occurs, the Delaunay
triangulation might change and one of our tricked facets might
disappear, leading to a change back to poiseuille controlled
permeability and possible instability in the solution of the linear
system.

If this is truly the problem, and our solution is to lock the positions
of the particles (w.r.t PFV), I don't think the DFNFlow scheme is truly
poroelastic since we aren't allowing the change of mechanical volume of
the pore to contribute to the pressure in the pore.

Yes Timos, the solution seems to be that we need a new way to relate
crack-flagged facets to cracks. Actually, this raises another concern
I've had, which is that the current method cannot trick facets near a
broken bond that does not line up with an edge on the triangulation. I
wonder if we could trick facets by looking for their intersection with a
broken bond, instead of looking for edges that are broken bonds. This
would be more or less an entire rewrite of DFNFlow and the intersection
algorithm might be expensive...and it may still not solve the
instability problems discussed in this thread :-(.

Bruno, do you see a solution to this that can be implemented solely on
Yade's side of things, without diving into CGAL's libraries? It seems to
me we could "force" CGAL to use the Delaunay triangulation once,
followed by a continuous use of the same vertex IDs for the same cells
during subsequent triangulations. I doubt this sort of functionality is
included in CGAL. If we need to dive into CGAL, do you see an easy
solution within CGAL or is it a likely headache? I only ask because I
know you have your head much deeper into the CGAL side of things than
me. I'm willing to implement the solution, but I want to make sure I am
headed down the right path before starting.

Best,

Robert

-- 
You received this question notification because your team yade-users is
an answer contact for Yade.