Geometry & Meshing#

PalaceToolkit uses Gmsh with its OpenCASCADE kernel to create parametric 3D finite-element meshes entirely from Python scripts.

The Entity abstraction#

Every geometric region — conductors, dielectrics, air boxes, ports — is represented as an Entity:

from palacetoolkit.mesh import Entity, run_meshing_pipeline

conductor = Entity(name="conductor", dim=3, mesh_order=0, tags=[1])
dielectric = Entity(name="dielectric", dim=3, mesh_order=1, tags=[2])
air = Entity(name="air", dim=3, mesh_order=2, tags=[3])

Parameter

Description

name

Becomes the physical group label that Palace references.

dim

Geometric dimension (3 = volume, 2 = surface, …).

mesh_order

Priority for the boolean pipeline (lower = higher priority).

tags

Gmsh geometry tags returned by gmsh.model.occ.* calls.

Boolean pipeline#

After defining entities, call run_meshing_pipeline to perform automatic priority-based cuts and fragmentation:

run_meshing_pipeline([conductor, dielectric, air])

The pipeline:

  1. Groups entities by dimension (3 → 0).

  2. Sorts within each dimension by mesh_order (ascending = higher priority).

  3. Cuts each entity against all previously processed entities in the same dimension (removeObject=True, removeTool=False).

  4. Fragments lower-dimensional entities against higher-dimensional ones so that boundary surfaces are shared.

  5. Assigns physical groups — volumes get their entity name; surfaces at material interfaces are auto-labelled (e.g. "conductor__dielectric").

Tip

The mesh_order controls which material wins at overlapping regions. Think of it as a z-index: small number = highest priority = never cut.

Mesh generation#

After the pipeline runs, generate the mesh and write it to disk:

import gmsh

gmsh.model.mesh.generate(3)
gmsh.write("my_model.msh")
gmsh.finalize()

Mesh topology verification#

Before running Palace, verify the mesh is valid:

from palacetoolkit.verify_topology import analyse_mesh

analyse_mesh("my_model.msh")

This checks that:

  • Every boundary element maps to an existing volume face (totality).

  • No two boundary elements map to the same face (injectivity).

Mesh visualisation#

from palacetoolkit.viz import view_mesh

view_mesh("my_model.msh")

view_mesh() uses PyVista’s interactive notebook backend when executed in Jupyter, so the mesh remains pan/zoom/rotate capable in rendered docs.