Degenerate 3D Elements with Gmsh mesh

Mesh generators, CAD programs, and other tools
Post Reply
mmccrea
Posts: 3
Joined: 09 Jul 2014, 00:11
Antispam: Yes

Degenerate 3D Elements with Gmsh mesh

Post by mmccrea »

Good day.

I have recently started to have some problems with using gmsh and Elmer to create field maps for use in Garfield++. Initially using Gmsh 2.6.1 with Elmer library version 7.0 (rev 6109M) under MacOsX 10.8.5 the meshing and solving worked. For larger meshes and more events I am trying to move to my groups simulation server.

My meshing and solver commands are:

> gmsh n3HeTarget.geo -3 -order 2
> ElmerGrid 14 2 n3HeTarget.msh -autoclean
> ElmerSolver n3HeTarget-400V.sif

When moving to the server it was required to update gmsh from 2.6.1 for 64bit support. I went to gmsh2.8.5. I also changed to the newest (at the time) Elmer, version 7 (rev7023).

Now using the same input files the meshing and elmergrid seem to work but the ElmerSolver now finds a large number of degenerate points in the mesh and gives error messages of the form:

ERROR:: ElementMetric: Degenerate 3D element: 191783
ElementMetric: DetG: -1.735E-18
ElementMetric: Node: 1 Coord: 9.754E-01 6.739E+00 1.270E+00
ElementMetric: Node: 2 Coord: 9.754E-01 6.439E+00 1.270E+00
ElementMetric: Node: 3 Coord: 1.474E+00 6.888E+00 1.233E+00
ElementMetric: Node: 4 Coord: 1.474E+00 6.289E+00 1.233E+00
ElementMetric: Node: 5 Coord: 9.754E-01 6.589E+00 1.270E+00
ElementMetric: Node: 6 Coord: 1.225E+00 6.664E+00 1.252E+00
ElementMetric: Node: 7 Coord: 1.225E+00 6.813E+00 1.252E+00
ElementMetric: Node: 8 Coord: 1.225E+00 6.514E+00 1.252E+00
ElementMetric: Node: 9 Coord: 1.225E+00 6.364E+00 1.252E+00
ElementMetric: Node: 10 Coord: 1.474E+00 6.589E+00 1.233E+00
ElementMetric: Node: 2 dCoord: 0.000E+00 -2.995E-01 0.000E+00
ElementMetric: Node: 3 dCoord: 4.987E-01 1.497E-01 -3.653E-02
ElementMetric: Node: 4 dCoord: 4.987E-01 -4.492E-01 -3.653E-02
ElementMetric: Node: 5 dCoord: 0.000E+00 -1.497E-01 0.000E+00
ElementMetric: Node: 6 dCoord: 2.493E-01 -7.487E-02 -1.827E-02
ElementMetric: Node: 7 dCoord: 2.493E-01 7.487E-02 -1.827E-02
ElementMetric: Node: 8 dCoord: 2.493E-01 -2.246E-01 -1.827E-02
ElementMetric: Node: 9 dCoord: 2.493E-01 -3.744E-01 -1.827E-02
ElementMetric: Node: 10 dCoord: 4.987E-01 -1.497E-01 -3.653E-02

it tries to complete the field solution and looks like it might have worked but the field is wrong or won't load in Garfield++. So while the first error message is given by ElmerSolver but the the .msh generated by gmsh2.8.5 and elmergrid is significantly reduced in size (file size and node count) compared to the original mesh made with gmsh2.6.1,

Opening either mesh in gmsh seems to look good, all of the physical volumes and the overall mesh volume look good, and similar.

Has anyone had similar problems?
or know if the .msh output format has changed in gmsh?

My .geo and attached as well as a terminal logs for each setup showing what was done.

.sif and auxillary geometry files are beyond the 3 file limit.

Mark McCrea
Attachments
n3HeTarget.geo
gmsh geometry file
(5.79 KiB) Downloaded 371 times
TerminalLog-gmsh285onserver-mesh-solve2015_03_05.txt
Terminal output log from server using new versions of gmsh and elmer where it doesn't work.
(221.19 KiB) Downloaded 353 times
TerminalLog-gmsh261onserver-mesh-solve2015_03_06.txt
Terminal output from macbook with older versions where it does work.
(201.95 KiB) Downloaded 350 times
mmccrea
Posts: 3
Joined: 09 Jul 2014, 00:11
Antispam: Yes

Re: Degenerate 3D Elements with Gmsh mesh

Post by mmccrea »

Good day.

Here are the remaining 3 files required to reproduce my setup for testing.

Mark McCrea
Attachments
n3HeTarget-400V.sif
(1.88 KiB) Downloaded 332 times
gf_cylinder.geo
(2.63 KiB) Downloaded 356 times
gf_rectangle.geo
(2.93 KiB) Downloaded 361 times
mzenker
Posts: 1999
Joined: 07 Dec 2009, 11:49
Location: Germany

Re: Degenerate 3D Elements with Gmsh mesh

Post by mzenker »

Hi,

I had similar problems already. In my cases, it normally helped to optimize the mesh in gmsh (Mesh->Optimize3D or Optimize3D(Netgen)).

HTH,

Matthias
mmccrea
Posts: 3
Joined: 09 Jul 2014, 00:11
Antispam: Yes

Re: Degenerate 3D Elements with Gmsh mesh

Post by mmccrea »

As mzenker suggested for this problem part of the solution was to add -optimize to the end of the gmsh command.

During my testing I had also left the mesh pretty coarse around the wires. When I got in touch with the gmsh mailing group with this question one of the developers found due to a change between after 2.7.1 they had changed some of their internal meshing parameters in gmsh so the coarse meshing I told the program to start with now failed to find a good solution and also would not optimize. I had to reduce the mesh size around my wires to get the optimization to work. My file size is now nearly twice as large (~200->~400megs) but the program works better now.

Further mesh fineness increases are probably in the future as some convergence issues still occur at certain points around the wires modeled a thin cylinders in my simulation.
Post Reply