Constructions in unstructured models

We are often making traditional layered feflow models with complicated structures, such as tunnels and caverns. If the geology is complicated and the structures are too, the layered approach becomes difficult and thus we are looking towards unstructured models. In this case the geology may be imported from a geological modelling software, but what about the structures ?. We usually start with a 3D autocad file, but as I understand it, feflow (tetgen) can only use PLCs to make the grid ?. Is this correct ?. If the structure is simple, its easy enough to make a PLC by hand, but when it's a long winding tunnel, this is not a viable option. Can anyone suggest a way of getting a 3D autocad file into feflow to be used for construction of an unstructured mesh. We would probably need to simplify the structure before using it for meshing and often there are many points in the Autocad 3D file that does not conform to PLC rules due to inaccuracies in the drawing. Any suggesions will be appreciated.

Regards Christian

Christian,

indeed this is one of the most complex steps when working with unstructured meshes. I cannot offer a general, easy and straightforward way so far (as we're dealing with extremely complex geometrical problems here), but I've used the following workflows in different combinations so far:

- re-work AutoCad (or shp) files in a software like Blender that (for simple to moderately complex cases) can detect intersections etc. and partly also resolve them automatically (remove duplicates, intersect polygons, re-distribute vertices, ...)
- use the vertices of the structures only for meshing rather than surfaces - but this may compromise the integrity of the structures, and makes parametrization more complicated
- develop my own pre-processing code for building PLCs (in my case for geology with vertical faults)

In general, TetGen can use PLCs along with 'add-ins', so points, lines, and non-closed triangulated surfaces work as well (also for re-meshing parts of a layered mesh - often a good option for technical structures in regional models). With respect to PLCs, FEFLOW does a number of simple corrections on its own (like duplicate polygons) before handing over geometry to TetGen, but not more complicated ones.

I hope this can help a bit - but as said, producing viable input geometries for unstructured meshing remains a challenging task.

Cheers,
Peter

Thanks for the suggestions.. :)