Re: [femlisp-user] Which Mesher?

[Nicolas Neuss]

Todd Pierce <address@hidden> writes:

> Hello, 
>
> I am developing a natural language processor which (hopefully) will
> have some FEM (maybe CFD?) capabilities.   Presently, I'm trying to
> figure out the best way to create the CAD models and mesh them
> completely from the command line.
>
> Obviously, pointing and clicking on things is not going to work since
> the CAD/mesh product will have to receive input (in its native
> language) from my program through a pipe or similar stream.
>
> I have had some luck doing this with GMSH.  There is an
> interactive.cpp program which gives GMSH a command line and meshes
> come out great.  Not sure if it's the best solution though.
>
> My question is, would the meshes generated (.msh files discussed
> here: http://www.geuz.org/pipermail/gmsh/2010/005329.html) be
> compatible with FEMLisp?

Hello Todd,

it depends on what type of elements GMSH uses for generating the FE
mesh.  For example, if the result consists only of tetrahedra (which I
would expect to be the easiest choice for 3D mesh generation), it should
be very straightforward to import into Femlisp.

Do you have a sample file of some mesh generated by GMSH?

> Is there a better way to approach creating
> a CAD model and meshing it from the command line?
>
> -Todd

I am not really experienced in CAD models.  Internally, a Femlisp domain
is a so-called "cell complex" where the basic cells are simplices and
simplex products mapped by arbitrary nonlinear mappings.  This is very
general and should be sufficient for modeling arbitrarily complicated
domains.  CAD descriptions are much more specialized, and it can be
quite some work to map those onto the Femlisp cells.

Nicolas