Importing a GMSH mesh into FEniCS

General
1

I am now trying to use FEniCSx/Dolfinx to solve a simple equation but I want to import a gmsh mesh which has boundaries etc. However, whenever I try this I get errors such as:

Traceback (most recent call last):
File “/home/fenics/HeatConduction/src/heat_conduction/test4.py”, line 60, in
msh = xdmf.read_mesh(name=“Grid”)
File “/usr/local/dolfinx-real/lib/python3.10/dist-packages/dolfinx/io/utils.py”, line 258, in read_mesh
cell_shape, cell_degree = super().read_cell_type(name, xpath)
RuntimeError: Cannot recognise cell type. Unknown value: mixed

I am confused to what I am doing wrong!

Note I have tried using Docker Dolfinx and conda but neither works (the later doesn’t even install fenics/dolfinx!). I am on an M3 MBP.

The code I am using is:

import gmsh
import meshio
from mpi4py import MPI
from dolfinx.io import XDMFFile
import dolfinx.fem as fem

def create_doughnut_mesh(filename, refinement_level):
    gmsh.initialize()
    gmsh.model.add("Doughnut")

    # Define the outer and inner circles
    outer_circle = gmsh.model.occ.addCircle(0, 0, 0, 1)
    inner_circle = gmsh.model.occ.addCircle(0, 0, 0, 0.5)

    # Create the plane surface between the circles
    outer_loop = gmsh.model.occ.addCurveLoop([outer_circle])
    inner_loop = gmsh.model.occ.addCurveLoop([inner_circle])
    surface = gmsh.model.occ.addPlaneSurface([outer_loop, inner_loop])

    # Synchronize the CAD kernel with the Gmsh model
    gmsh.model.occ.synchronize()

    # Ensure no quadrilaterals are created
    gmsh.option.setNumber("Mesh.CharacteristicLengthFactor", 1.0 / refinement_level)

    # Generate the mesh
    gmsh.model.mesh.generate(2)

    # Ensure the mesh contains only triangles
    element_types, element_tags, node_tags = gmsh.model.mesh.getElements()
    if len(element_tags) == 0 or len(element_tags[0]) == 0:
        raise RuntimeError("Mesh generation failed: no elements found.")
    if any(element_type != 2 for element_type in element_types):  # 2 corresponds to triangles in Gmsh
        print(element_types)

    # Get the boundary entities
    outer_boundary = gmsh.model.getBoundary([(2, surface)], oriented=False, recursive=False)
    inner_boundary = gmsh.model.getBoundary([(2, surface)], oriented=False, recursive=False)

    # Create physical groups for the boundary edges
    outer_boundary_group = gmsh.model.addPhysicalGroup(1, [edge[1] for edge in outer_boundary], tag=1)
    inner_boundary_group = gmsh.model.addPhysicalGroup(1, [edge[1] for edge in inner_boundary], tag=2)
    gmsh.model.setPhysicalName(1, outer_boundary_group, "OuterBoundary")
    gmsh.model.setPhysicalName(1, inner_boundary_group, "InnerBoundary")

    # Optionally, save the mesh to a file
    gmsh.write(filename)

    gmsh.finalize()

# Example usage
create_doughnut_mesh("doughnut.msh", refinement_level=10)

# Convert the Gmsh mesh to XDMF format
mesh = meshio.read("doughnut.msh")
meshio.write("doughnut_filtered.xdmf", mesh)

# Load the mesh into FEniCSx
with XDMFFile(MPI.COMM_WORLD, "doughnut_filtered.xdmf", "r") as xdmf:
    msh = xdmf.read_mesh(name="Grid")

# Read the filtered mesh from the XDMF file
filtered_mesh = meshio.read("doughnut_filtered.xdmf")

# Extract boundary edges
boundary_edges = []
for cell_block in filtered_mesh.cells:
    if cell_block.type == "triangle":
        for triangle in cell_block.data:
            for i in range(3):
                edge = tuple(sorted([triangle[i], triangle[(i + 1) % 3]]))
                if edge not in boundary_edges:
                    boundary_edges.append(edge)

# Print the boundary edges
print("Boundary edges:")
for edge in boundary_edges:
    print(edge)

V = fem.FunctionSpace(msh, ("CG", 1))
2

See for instance: Can dolfinx read a GMSH file containing 1D mesh of an ellipse? - #2 by dokken
or use the built in GMSH API functionality: dolfinx.io.gmshio — DOLFINx 0.8.0 documentation
which work directly on the gmsh model rather than storing it into an intermediate format…

This seems unrelated to the current post, and you have not provided the steps you followed and your error trace. Steps for installing with conda is outlined at for instance: https://fenicsproject.org/download/