magnet.mesh.Mesh

class magnet.mesh.Mesh(adjacency: csr_matrix, coords: ndarray, volumes: ndarray, physical_groups: ndarray | None = None)

Bases: object

Graph data that describes a mesh.

Each cell of the mesh corresponds to a node of the graph: 2 nodes are connected by an edge if the corresponding cells are adjacent in the mesh. Each node is also characterized by its node features: coordinates and a area/volumes of the respective cell. This information is held in 3 corresponding attributes (Adjacency, Coords, Volumes). Heterogeneity is described by an additional node feature, Physical_Groups.

Used as interface for running agglomeration models and training GNNs. It does not contain the full data needed to describe the mesh (that is reserved to the AgglomerableMeshHeterogeneous class).

Parameters:

• adjacency (sparse.csr_matrix of np.uint8) – Adjacency matrix describing the mesh.

• coords (np.ndarray of float) – Centroid coordinates of each cell of the mesh.

• volumes (np.ndarray of float) – Volumes (or areas, if the mesh is 2D) of the cells.

• physical_groups (np.ndarray of float) – Physical group of each cell. Values should be between 0 and 1.

Variables:

• dim (int) – Spatial dimensions of the mesh (2 or 3).

• num_cells (int) – Number of cells in the mesh.

• Adjacency (sparse.csr_matrix of np.uint8) – Adjacency matrix of shape (num_cells, num_cells) describing the mesh. Adjacency[i, j]=1 if cells i and j are adjacent, 0 otherwise.

• Coords (np.ndarray of float) – Array with shape (num_cells, dim) containing the centroid coordinates of each cell of the mesh.

• Volumes (np.ndarray of float) – Array with shape (num_cells, 1) containing the volumes (or areas, if the mesh is 2D) of the cells.

• Physical_Groups (np.ndarray of float, optional) – Array with shape (num_cells,1) containing the physical group of each cell. By default, set all physical groups to 0, i.e. the mesh is considered homogeneous.

See also

AggMesh

Agglomerable version.

Constructor

__init__(adjacency: csr_matrix, coords: ndarray, volumes: ndarray, physical_groups: ndarray | None = None) → None

Methods

__init__(adjacency, coords, volumes[, ...])
subgraph(node_ids)	Extract submesh corresponding to given element ids.
visualize_as_graph([classes, ...])	Visualize the graph representing the mesh.

__init__(adjacency: csr_matrix, coords: ndarray, volumes: ndarray, physical_groups: ndarray | None = None) → None

subgraph(node_ids: Iterable[int]) → Mesh

Extract submesh corresponding to given element ids.

Parameters:

node_ids (Iterable of int) – Iterable containing the ids of the elements that will bellong to the new submesh.

Returns:

The new submesh.

Return type:

Mesh

visualize_as_graph(classes: List[ndarray[int]] | None = None, view_phys_groups: bool = False, palette: Iterable[str] | None = None, edge_color: str = 'tab:gray', edge_width: float = 1, node_size: Iterable[float] | None = None) → None

Visualize the graph representing the mesh.

If a partition of the graph defined by classes is provided, color each portion differently. Alternatively, using view_phys_groups=True, each physicallly heterogeneoous part of the mesh is colored differently.

Parameters:

• classes (ClassList, optional) – Graph partition (e.g. coming from an agglomeration model). If provided, color each set differently.

• view_phys_groups (bool, optional) – If True, color nodes based on physical groups; in that case, classes parameter will be ignored. Default id False.

• palette (str or Iterable[str], optional) – Single color for graph nodes, or color palette to be used in conjunction with classes or view_phys_groups. By default, uses matplotlib default color palette.

• node_size (Iterable[float], optional) – Size of each node drawn in the graph. If a single value is provided, draw all nodes with that size. By default, node dimensions scale with the corresponding element area/volume.

• edge_color (str, optional) – Color of the graph edges. Default is tab:gray.

• edge_width (float, optional) – Width of the graph edges.

Return type:

None