[Gmsh] Zero volume elements in Delaunay 2d triangulation

Jose Paulo Moitinho de Almeida moitinho at civil.ist.utl.pt
Wed Apr 2 10:00:47 CEST 2008


On Tuesday 01 April 2008, Jean-Fran�çois Remacle wrote:
>.....
>
> > I got the attached screenshot with the following command
> > gmsh ref.geo -algo frontal -bgm ref.xxx2 -rand 1.e-11
> > which produces "wrong elements" and was just wondering what is the cause
> > of the "missing edges".
>
>  We use robust predicates for doing geometrical operations
> on the 2D mesh, so we should certainly be able to do better.
> I'll have a look at that whan I have time.
>

I was looking at the numbers, the highest mesh gradients have numbers like 
this
ST(  5.0000098e-01,  8.0000078e-01, 0,  
       4.9999966e-01,  8.0000118e-01, 0,  
       5.0000000e-01,  8.0000000e-01, 0)
{  4.8583228e-07,   4.8583228e-07,   7.7381012e-09};
i.e. we want to jump from 1e-7 to 1e-9 in a distance of about 1e-6, which 
isn't exactly a smooth transition. We will have a look at our refinement 
criteria.

> For the "graphical issue", the strange behavior is principally due to
> the way
> we use graphics : the zoom exceeds the resolution of the viewport. There
> again,
> we should be able to do better. So, what you see is NOT what you get.
>

I hope it may be an help. The results that we are getting using your code are 
already very good.

Regards

ZP