Constructing FEniCS meshes for structured geometries by tiling the domain with
some representative tile.
The master branch is compatible with FEniCS 2017.2.0 (python2, no pybind11)
and FEniCS 2018.+ (python3, pybind11).
Put this directory on python path, e.g. by running source setup.rc. For
checking the installation execute python -m unittest discover test or
python3 -m unittest discover test
In order to get the mesh for EMI experiments navigate to ./tiles. Make
sure that your Gmsh version is at least 3.0.6 but less than 4.+. In the following
the tile tile_1_narrow_GMSH306.geo is repeated 3 times in x and 4 times in y
direction to create the mesh.
- Get the mesh for the tile:
gmsh -3 -clscale 0.3 tile_1_narrow_GMSH306.geo - Wrap the mesh with data to H5:
python msh_convert.py tile_1_narrow_GMSH306.msh - Tile the domain:
python tiled_mesh.py tile_1_narrow_GMSH306.h5 -m 3 -n 4 -scale_x 1E-3
Where scale_x argument makes the units of mesh to be milimiters. Further, python is
either python2 or python3. Note that Gmsh version should match that of the tile.
(The .geo files have hardcoded mapping for periodic entities and their numbering seems
to change between versions.). Also note that 1.-3. can only run in serial.
There is an optional step 4 which removes those (heart) cells of the mesh that touch the mesh boundary
mpirun -np 3 python3 deactive_cells.py tile_1_narrow_GMSH306_3_3.h5 -m 3 -n 3
Instead of specifying the cell in the geo file we can use Gmsh Python API directly
- Get the mesh for the tile:
python3 tile_1_narrow.py -clscale 0.3 -format msh2 - Wrap the mesh with data to H5:
python msh_convert.py tile_1_narrow.msh - Tile the domain:
python tiled_mesh.py tile_1_narrow.h5 -m 3 -n 4 -scale_x 1E-3
Note that -format msh2 flag is necessary for step 2 to work. For conversion betweem
msh and dolfin compatible formats you could instead use meshio.
This package is CI tested against FEniCS packages for ubuntu 16.04 LTS
