[Gmsh] crash on 3D meshing (AW: Anisiotropic mesh in stacked layers)

[Zenker, Dr. Matthias]

Hi Christophe,

thank you for the quick reply!
I have continued to experiment on my side. It seems that mmg3d can cope with internal surfaces if the mesh is not too fine. If I use h2/20 in my example (with the interface), gmsh doesn’t crash. It also seems that a larger absolute mesh width does not improve things: If I multiply all lengths by 10, I still get the crash, even with h2/20. BTW gmsh takes more time to do the 2D mesh with all lengths multiplied by 10 even though the number of triangles should be almost unchanged.
Could it be that the sudden change in mesh width in z direction is a cause of the problem? I might change my field definitions in order to have a more gradual change of mesh width – do you think this would help?

Matthias

Von: Christophe Geuzaine [mailto:cgeuzaine at ulg.ac.be]
Gesendet: Donnerstag, 11. Juli 2013 22:58
An: Zenker, Dr. Matthias
Cc: gmsh at geuz.org Geuzaine; Jean François Remacle
Betreff: Re: crash on 3D meshing (AW: Anisiotropic mesh in stacked layers)


Hi Matthias,

There are two separate problems:

1) Currently MMG3D does not like having internal surfaces in the volume. Attached is a file with a single volume, which works ok.

@JF : have you tried MMG3D in such cases with internal surfaces?

2) "h2 / 100" pushes the initial tetrahedralization to the limit; h2/90 works fine over here.

All of this is still very experimental, though. It all needs to be tested and documented.

Thanks for the feedback!

Christophe


On 11 Jul 2013, at 17:29, "Zenker, Dr. Matthias" <Matthias.Zenker at erbe-med.com<mailto:Matthias.Zenker at erbe-med.com>> wrote:

> I forgot to say that the crash happens under Windows 7 with gmsh 2.8.0 64bit and with 2.7.1 32bit and 64bit.
>
> Matthias
>
> _____________________________________________
> Von: Zenker, Dr. Matthias
> Gesendet: Donnerstag, 11. Juli 2013 16:57
> An: 'gmsh at geuz.org'
> Cc: 'Christophe Geuzaine [cgeuzaine at ulg.ac.be] (cgeuzaine at ulg.ac.be<mailto:cgeuzaine at ulg.ac.be>)'
> Betreff: crash on 3D meshing (AW: Anisiotropic mesh in stacked layers)
>
>
> Hi,
>
> now that I have successfully created a anisotropic mesh width field for my geometry, I get a crash when I do the 3D meshing. 2D meshing works without problem. I have the impression that the desired mesh width may be too small. Is this so, and what is the smallest mesh width gmsh can cope with?
>
> Thanks for an answer, this time…
>
> Matthias
>
> < Datei: layers.geo >>
> _____________________________________________
> Von: Zenker, Dr. Matthias
> Gesendet: Donnerstag, 11. Juli 2013 16:41
> An: 'gmsh at geuz.org'
> Cc: 'Christophe Geuzaine [cgeuzaine at ulg.ac.be] (cgeuzaine at ulg.ac.be<mailto:cgeuzaine at ulg.ac.be>)'
> Betreff: AW: Anisiotropic mesh in stacked layers
>
>
> Hi again,
>
> I have got it to work, but I still don’t understand why. It seems that the fields in MathEvalAniso have to be 1/(mesh_width)^2. But why?
> I have read the relevant part of the gmsh manual, but didn’t find the answer there.
> A short explanation and/or a pointer to more info would be highly appreciated.
>
> Matthias
>
> _____________________________________________
> Von: Zenker, Dr. Matthias
> Gesendet: Mittwoch, 10. Juli 2013 17:09
> An: gmsh at geuz.org<mailto:gmsh at geuz.org>
> Cc: Christophe Geuzaine [cgeuzaine at ulg.ac.be] (cgeuzaine at ulg.ac.be<mailto:cgeuzaine at ulg.ac.be>)
> Betreff: Anisiotropic mesh in stacked layers
>
>
> Hi,
>
> I want to mesh a geometry with layers stacked in z direction, where each layer should have a mesh width in the xy plane, and a different one in z direction. I have tried to combine the Box, Min and MathEval Aniso fields to achieve this (see attachment). But the mesh widths are not as I expect, in particular in layer 2 in z direction it should be 1/20 its height. What do I do wrong?
> How is the mesh width derived from the m11, m22 and m33 fields in MathEvalAniso?
>
> Thanks for some explanations and hints,
>
> Matthias
>
> < Datei: layers.geo >>
>
>
>
> _____________________________________________________________________
> ERBE Elektromedizin GmbH
> Firmensitz: 72072 Tuebingen
> Geschaeftsfuehrer: Christian O. Erbe, Reiner Thede
> Registergericht: Stuttgart HRB 380137
>
> <layers.geo>

--
Prof. Christophe Geuzaine
University of Liege, Electrical Engineering and Computer Science
http://www.montefiore.ulg.ac.be/~geuzaine




_____________________________________________________________________
ERBE Elektromedizin GmbH
Firmensitz: 72072 Tuebingen
Geschaeftsfuehrer: Christian O. Erbe, Reiner Thede
Registergericht: Stuttgart HRB 380137

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.geuz.org/pipermail/gmsh/attachments/20130712/f8b740d3/attachment.html>