[Gmsh] Interior Element Growth

Johannes Wimmer johannes.wimmer at rwth-aachen.de
Tue Jun 23 12:03:59 CEST 2015


Hello again,
by now, I find that background meshes might work for the problem I have.

My workflow right now is as follows:

I mesh the volume in which I want high mesh density first and save the 
*.msh-file. Then, I convert this to *.pos format via:

Merge "highRes.msh"; Save "highRes.pos";

I got this from here: http://www.geuz.org/pipermail/gmsh/2013/008422.html

Then, I simply run: gmsh complete.geo -bgm highRes.pos

I've provided an example here: 
https://gigamove.rz.rwth-aachen.de/d/id/uabNjRNEwkD4Qw

I am encountering several problems now:
1) If there is a Physical Volume defined, only volume elements (I'm 
using tets) are present in the *.msh-file. This leads to non compatible 
meshes at the interface. It would be trivial to remove the definition of 
a Physical Volume from my input file, but is it possible to specify what 
is saved in the *.msh-file?

2) What is saved in the *.pos-file is element type, coordinates and the 
number of the elementary geometrical entity to which the element 
belongs. That is, 1 15 2 0 1 1 in *.msh becomes 
SL(0,0,0,0.0746901,0,0){17,17}; in *.pos.  I want the last part to be 
the target element size. Right now, I am simply replacing it using e.g. 
sed. Can GMSH do this for me.

3) The most severe issue for me is that it looks like GMSH does not 
always stick to the highRes mesh, but puts a coarse mesh where it should 
not. I have no clue how to address this. Maybe I am doing something 
wrong when I replace the number of the elementary geometrical entity 
with the target mesh size? It worked for one simple example, while more 
complex ones fail.

4) My geometries will become a lot larger. It says in the manual about 
*.pos: However this “parsed format” is read by Gmsh’s script parser, 
which makes it inefficient if there are many elements in the dataset. 
Also, there is no connectivity information in parsed views and all the 
elements are independent (all fields can be discontinuous), so a lot of 
information can be duplicated. For large datasets, you should thus use 
the mesh-based post-processing file format described in File formats, or 
use one of the standard formats like MED.

However, it also says that you can only use *.pos for background meshes. 
Is there some other option?

Cheers,
Johannes

-- 
Dipl.-Ing. Johannes Wimmer
RWTH Aachen
Institut für Angewandte Mechanik
Mies-van-der-Rohe-Str.1
D-52074 Aachen

Tel.: +49 (0)241 80 25013
Fax : +49 (0)241 80 22001
www.ifam.rwth-aachen.de