Meshing subsets of segments - brendanjmeade/celeri GitHub Wiki

Segments tagged with create_ribbon_mesh > 0 can be meshed into continuous networks of triangular dislocation elements, termed ribbon meshes, using the notebook segment_meshing.ipynb. This flag can be specified within a segment file, and the meshing takes place after the block closure routine so that the segments can be properly ordered.

First, the bottom coordinates of each segment are found, reckoning from the endpoints in a direction perpendicular to strike by the horizontal projection of the fault width; this is zero for vertical segments. Along a contiguous subset of segments, the bottom coordinates are averaged so that a closed loop of top and bottom vertices can be constructed. This means that, when adjacent segments differ in dip and/or locking depth, the originally specified segment properties do not strictly propagate into the TDE mesh. Geometric consequences include trapezoidal faults (where adjacent segments have different locking depths) and modification of specified dip. Creating shorter segments, including through repeated splitting to retain consistent properties, can reduce the impact of the bottom coordinate averaging.

Meshing is carried out using Gmsh, and specifically using several features. First, each segment (with averaged bottom coordinates) is specified as its own quadrilateral surface, but these surfaces are merged to create a compound surface (gmsh.model.mesh.setCompound). The original surfaces are generated using gmsh.model.geo.addSurfaceFilling rather than gmsh.model.geo.addPlaneSurface, which produced nonplanar surfaces in testing.

Meshing is carried out using geocentric Cartesian coordinates, but the TDE nodes are converted back to longitude, latitude, depth (km) coordinates before being written to a .msh file. A unique .msh file is written for each set of contiguous segments as (segment_file_name)_ribbonmesh1.msh, (segment_file_name)_ribbonmesh2.msh, etc. using the same folder as any originally specified meshes. We start with ..._ribbonmesh1.msh rather than ..._ribbonmesh0.msh so that the numerical index can correspond to an updated create_ribbon_mesh flag in the segment file.

In addition to generating the meshes, the ribbon meshing routine updates several Dicts and DataFrames, which are written to updated files for use in a subsequent celeri run:

• In Segment, patch_file_name is set to be a new integer corresponding to the ribbon mesh index (corrected for number of meshes in the original .mshp file), and patch_flag is set to 1. This allows the locking depth of the segment to be set to zero. create_ribbon_mesh is set to zero so that meshes are not recreated.
• The .mshp file is updated with new entries identifying the newly created ribbon meshes. These are added after any existing meshes and the listing order corresponds to the values in segment.patch_file_name.
• A new .command file is written, updated with the filenames of the new segment.csv and mesh parameter .json files. These are placed in the same folder as the original files, with _ribbonmesh appended.