Currently I am working on numerical modeling of 2-dimensional magnetotelluric data using finite element method. To generate a forward model, I discretize my model area (200000,100000) meters using a partial differential equation toolbox given below:
gd=[2 4 -100000 100000 100000 -100000 0 0 -100000 -100000];
[dl,bt]=decsg(gd');
wgeom(dl,'vrg');
[p e t]=initmesh('vrg');
figure
pdeplot(p,e,t);
xlabel('Distance(m)','fontsize',12);
ylabel('Depth(m)','fontsize',12);
axis image;
I want to generate model area which element size will increase with depth or with which depends on conductivity distribution of an element.
Please look at the images in the attachment .
You have mesh tools allowing you to specify mesh seed lengths with respect to the area. What you need is to decompose your initial domain as function of the mesh characteristic length desired.
A very basic idea to avoid direct parametered meshing is for example is to coarsely mesh your full domain, and then refine the interiors of specific areas down to the desired size. This way you can use element detection methods to specify the areas to be refined.
Meshing softwares like gmsh (www.gmsh.org) allow you to do that. If you are interested in integrated calls from MATLAB, you can check out the finite element toolbox for MATLAB openFEM (http://www.sdtools.com/openfem/) that provides such features (see fe_gmsh).
There are multiple ways to go for adaptive meshing. You can start coarse and adaptively refine based on a criterion (e.g. red/green refinement). An other method is to distribute nodes according to some density function and create an (unstructured) triangulation through Delaunay triangulation. There are multiple mesh generation toolboxes out there...
I think you need to use skin depth concepts to set up an auto-mesh routine. See Vozoff, 1972 or the attached text chapter.
- Badges
- Science topic