Closed baagaard-usgs closed 3 weeks ago
If you use knepley/fix-orientation-input
in PETSc and baagaard-usgs/fix-cohesive-cells-orientation
in PyLith, then the orientations are consistent
@baagaard-usgs I confirm that this also fixes the solve for degree 2 subduction-2d
Describe the bug
Cohesive cells are not forced to have a consistent orientation. In many simple meshes (for example, all of our full-scale tests), the cohesive cells have consistent orientations. In more complex meshes, the orientation of a few cohesive cells is likely to be wrong. This creates errors in the slip, traction change, and stresses.
To Reproduce
PETSc branch:
knepley/pylith
PyLith fork:https://github.com/baagaard-usgs/pylith.git
PyLith branch:fix-cohesive-cells-orientation
Steps to reproduce the behavior. Dump the
fault_normal_dir
vertex field inoutput/step01_slip-west_branch_info.h5
.BAD (normal dir should not flip)
Attempted fix
This generates an error: