Re: Calculations on data on triangular mesh.

[Richard]

On 25/10/2012 18:57, Juan Pablo Carbajal wrote:
> On Thu, Oct 25, 2012 at 4:25 PM, Richard Crozier <address@hidden> wrote:
> > Hi,
> >
> > I want to do some post-processing of data generated using a 2D finite
> > element code. The code produces the nodal quantities of the solution on a
> > delauney triangulation. The data I have in the solution are the vertex
> > locations, a map of these vertices for the triangles and values of the
> > solution at the vertices.
> >
> > What I would like to know is, is there a library of functions for
> > efficiently working with this kind of data? The kind of operations I am
> > interested in are interpolation at arbitrary points, integration over
> > regions of the triangulation, getting triangle areas, centroids etc etc. I
> > would like to avoid writing all this code from scratch.
> >
> > I have looked around for this on Octave Forge and so on, but haven't come
> > across anything appropriate yet.
> >
> > The code in question by the way is freely available and hosted here:
> >
> > https://sourced.ecdf.ed.ac.uk/projects/see/xfemm
> >
> > It is a mex interface I have created to the core C++ magnetics solver from
> > this program:
> >
> > http://www.femm.info/wiki/HomePage
> >
> >
> >
> >
> > --
> > View this message in context: 
> > http://octave.1599824.n4.nabble.com/Calculations-on-data-on-triangular-mesh-tp4645732.html
> > Sent from the Octave - General mailing list archive at Nabble.com.
> > _______________________________________________
> > Help-octave mailing list
> > address@hidden
> > https://mailman.cae.wisc.edu/listinfo/help-octave
> Hi Richard,
>
> Thank you for your link. Are you in contact with the developer of
> FEMM. I also use it but since it is not licensed under a
> GPL-compatible license we cannot add an interface in Octave Forge.
>
> You may find the packages msh, fpl and geometry relatively useful.
> There is many functionality missing and it would be great if you
> develop anything missing that you share it we us.
>
> To interpolate in a triangular mesh you could build the trick with
> griddata, but the best is to write your own bilinear interpolation (if
> that is what you want) on the triangles. I have code but is not
> documented. I can share if you want, but I do not promise good quality
> :D. Indeed... I may wrote a function to interpolate on a triangular
> mesh after the 29th...
>
> I am forwarding Carlo who is the one actively developing most of the FEM tools.

Juan,

I am in contact occasionally with the FEMM author, but he was not 
interested in my proposal which was to let me use my modified version of 
the magnetics solver code under an LGPL licence. I don't know if the 
FEMM author would be any more interested in the full GPL or another 
licence. In any case it uses Triangle (made by Jonathan Shrewchek) which 
is provided under similar terms to FEMM.

If you use FEMM, you might find my code useful, I have made a function 
(writefemmfile.m) which can output a file in FEMM's input format given a 
problem definition stored as a normal Octave/Matlab structure. The FEMM 
solver C++ code can also be compiled as a standalone executable as well 
as a mexfunction. The mexfunction works with Octave, too. My 'interface' 
mexfunction though really just calls the FEMM code with a string 
containing the name of .fem file. I have replaced all the Windows-only 
MFC code with standard C++ so it can also be compiled on linux, and 
doesn't require Wine. I have also created a lot of helper functions in 
m-code for creating the problem structure (e.g. addnode_mfemm.m, 
addsegment_mfemm.m and similar). These are analagous to the existing 
OctaveFEMM interface code.

Note this is only the magnetics solver though.

I want to implement all the post-processing capabilities of FEMM, but in 
Matlab and Octave compatible code. All mcode I have written for 
pre-processing (problem definition) and post-processing is released 
under the Apache V2.0 licence. The modified FEMM sources are provided 
under an Aladdin Free Public Licence.

At some point in the long term future I can consider replacing the 
actual solver code as well with something with a less restrictive 
licence. The alternative is to write a magnetics solver for FAESOR which 
is a 3D GPL finite element package for Matlab (haven't tested this 
extensively in Octave, but I'm pretty sure it works, apart from some 
visualisation code.). You can find FAESOR here:

http://hogwarts.ucsd.edu/~pkrysl/faesor/faesor_publish.html

In case you are interested. If you would like to help make the 
post-processing code for my FEMM interface, you, or anyone else are more 
than welcome!

Richard








--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.