[Gmsh] Aborting 3d meshing when points are (relative) too close to each other

Tilo Winter tilo.winter at uni-rostock.de
Thu Aug 29 01:29:50 CEST 2013


Hello together,

I have a problem with the „weird point“ message I get when I try to mesh (gmsh -3 ....geo) my generated .geo-file. 

In the context of my final thesis I try to generate a new mesh for the pitzDaily-tutorial of OpenFOAM with Gmsh, based on streamline data, which I extracted from ParaView as a .csv-file. When I generate too many streamlines or probably better to close points, which describe the streamlines, I get the error-message you can see below.

Info    : Running 'gmsh -3 pitzDaily_input.geo' [Gmsh 2.8.1, 1 node, max. 1 thread]
Info    : Started on Thu Aug 29 00:44:39 2013
Info    : Reading 'pitzDaily_input.geo'...
...
Error   : Weird point 110641664 in Coherence
Error   : Weird point 31727 in Coherence
Error   : Weird point 110781440 in Coherence
Error   : Weird point 0 in Coherence
Error   : Weird point 31831 in Coherence
Error   : Weird point 1 in Coherence
Error   : Weird point 111313664 in Coherence
Error   : Weird point 0 in Coherence
Error   : Weird point 105480144 in Coherence
Error   : Weird point 3 in Coherence
Error   : Weird point 105641504 in Coherence
Error   : Weird curve -478 in Coherence
Error   : Weird curve 1102 in Coherence
Error   : Weird curve -483 in Coherence
Error   : Weird curve 2490 in Coherence
Error   : Weird curve 3120 in Coherence
Error   : Weird curve 3122 in Coherence
Error   : Weird curve 3128 in Coherence
Error   : Weird curve 3741 in Coherence
…


I read in this thread (http://geuz.org/pipermail/gmsh/2007/002822.html) that it could be, that some points are too close together so that the 3D mesh algorithms become unhappy. (exactly here: http://geuz.org/pipermail/gmsh/2007/002831.html)

One imaginable solution was to scale up the whole OpenFOAM-case with transformPoints -scale '(…)' to generate larger absolut distances between the points, mesh the new .geo-file and to rescale the whole new OpenFOAM case back, but it doesn't work as I supposed. So I think that Gmsh must considers the relative distance between the points, not the absolut values!? Is that right?

Now my question is, if there is any option to solve this problem and to generate meshes with points closer to each other respectively with data of arbitrary closed streamlines(-data)?

Attached you find two screenshots of my problem with points probably (relative) to close to each other.

When I generated a few streamlines less respectively points, that are further away from each other, there is no problem. I also was able to increase the number of cells in each direction and everything worked.

I use Gmsh 2.8.1 64 bit on Ubuntu 12.04.

Thanks in advance for any help

Best regards,

Tilo
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 30pts_Error.jpg
Type: image/jpeg
Size: 721078 bytes
Desc: 30pts_Error.jpg
URL: <http://www.geuz.org/pipermail/gmsh/attachments/20130828/e2f813cd/attachment.jpg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 30pts_Error_detail.jpg
Type: image/jpeg
Size: 308070 bytes
Desc: 30pts_Error_detail.jpg
URL: <http://www.geuz.org/pipermail/gmsh/attachments/20130828/e2f813cd/attachment-0001.jpg>