[Gmsh] importing .step files with multiple parts whose surfacestouch

[Vandevender, Brent A]

Thank you for your reply Matthias.  I specifically had you and Mark van Doesburg in mind when I posed the question.  I have followed your posts in the archives and hoped that there was some progress.  I have only recently realized the gravity of my problem and start to appreciate the astronomically high price of the commercial electricity and magnetism FEM package I am trying not to have to buy (Maxwell3D).  It seems I will take some effort to solve the problem myself, as you have.  The best solution, I think,  is to fix the 2D surface meshes created by gmsh with a separate self-written code and re-import to gmsh as you describe in option 2.  I can imagine how the algorithm will work, but also that it will take the age of the universe to complete for a realistic geometry.  My intention was to be a mere consumer of finite-element methods, and I don't know where I will find the time to become a producer.


On 3/31/11 4:20 AM, "Zenker, Dr. Matthias" <Matthias.Zenker at erbe-med.com> wrote:

Hi Brent,

I have had (and posted) this problem, and still have it.
What has been done AFAIK is that a functionality which should in principle do what we need has been implemented in gmsh, developped by Mark van Doesburg. But it works only in some cases (e.g. plane surfaces). If you want to try it, check "Tools->Options->Geometry->Cut and merge faces (experimental)", reload your geometry and see what happens.
I had thought for a while that the geometry healing functionality of Netgen could solve the problem, but it does not, at least not in 3D.
I am actually working on a solution of the problem, but I don't know yet what it will be. I see two possibilities:

1. Switch to another preprocessor. I evaluate GiD at the moment, which is not free, but not so expensive either. As far as i have found out so far, it can collapse the surfaces. But I still have some technical problems, i.e. GiD crashes sometimes for complicated geometries. If this problem cannot be solved by the GiD people, I have to opt for solution 2.

2. Program something myself. I have actually started to write a mesh processor. The idea is to take the 2D mesh (e.g. generated by gmsh), identify touching surfaces, merge them, reload the cleaned up mesh in gmsh, re-define the volumes, and generate the 3D mesh. For the moment, my program just reads and orders the mesh. I will continue to work on it as long as the GiD problem is not solved.

That's the status from my point of view.

Matthias



--
Brent A VanDevender, PhD
Pacific Northwest National Laboratory
Radiation Detection and Nuclear Sciences Group

email:brent.vandevender at pnl.gov
phone:509-371-7982