pymesh

Chromatography meshing software built on top of gmsh.

A rewrite of genmesh because python is a bit more flexible and concise for what is essentially a wrapper around gmsh.

Plus a rewrite helps me reorganize the software at an architectural level.

Install

The best idea is to just use conda, create your environment, install necessary dependencies, install pip in conda, and then do pip install .

If gmsh is built from source, ensure that - PYTHONPATH points to $GMSH_ROOT/lib (or wherever the gmsh.py file is) - LD_LIBRARY_PATH points to $GMSH_ROOT/lib (or wherever the gmsh{.so,.a} files are)

NOTE: I use some weird git_version() function in setup.py. You might wanna change that.

Usage

The script in bin/ should be available in $PATH after the install. It is currently called mesh. Given the appropriate input file in yaml format, run: mesh input.yaml

Dummy yaml

packedbed:
  packing_file:
    filename: packing.xyzd
    dataformat: <d
  nbeads: 4
  scaling_factor: 1.0
  # auto_translate: True
  particles:
    scaling_factor: 0.9997
    # modification: bridge | cut 
    # relative_bridge_radius: ...
container:
  shape: cylinder
  size: [ 0.0, 0.0, -0.5, 0.0, 0.0, 4.0, 0.5 ]
  # size: [-2, -2, 0, 4, 4, 4]
  # periodicity: 
  # linked: True
  # stack_method: planecut
  # inlet_length: 0.0
  # outlet_length: 0.0
mesh:
  # size: 0.15
  size_method: field
  field:
    threshold:
      size_in: 0.06
      size_out: 0.14
      rad_min_factor: 0.4
      rad_max_factor: 0.6
  algorithm: 5
  algorithm3D: 10
  generate: 2
output:
  filename: mesh.vtk
gmsh:
  General.Verbosity: 5
  Geometry.OCCParallel: 1
  Mesh.MaxNumThreads1D: 8
  Mesh.MaxNumThreads2D: 8
  Mesh.MaxNumThreads3D: 8
  Mesh.ScalingFactor: 0.0001
general:
  fragment: True

Notes

• For shape: cyl, size: [x, y, z, dx, dy, dz, r]
• For shape: box, size: [x, y, z, dx, dy, dz]
• If mesh.field.threshold.size_in and mesh.field.threshold.size_out are not given, they default to mesh.size
• Set general.fragment to False to run a quick mesh and manual visual check for correct dimensions and intersecting volumes.
  • Best with mesh.generate set to 2
  • Be aware that this breaks physical groups, matching periodic surfaces etc

Features roadmap

• [DONE] Timestamp in log filenames
• [DONE] cylinder container shape
• [TASK] porosity manipulation by bead addition/deletion
• [TASK] Documentation
• [NEXT] bridges, cuts
• [TASK] Auto container sizing?
• [TASK] Move to numpy arrays?
• [TASK] check out numba njit
• [TASK] Parallelize
• [TASK] Profile stack_all
• [TASK] Debug mode
• [TASK] Add JSON support
• [TASK] Consider ruamel_yaml
• [TASK] Geometry save and load: stashed
• [TASK] Improve memory usage
  • separate_bounding_surfaces()
• [TASK] Print out information about the column:
  • length or bed and column
  • porosity of bed and column
  • Particle volumes
  • Particle size distribution: mean

Known issues:

• [CRIT] setting gmsh.General.NumThreads generates degenerate element surfaces. It's a gmsh issue.
  • observed when used with periodicity and linked mesh
  • In paraview, extract surfaces and then clip to see the degenerate surfaces
• OpenCASCADE fragment operation doesn't preserve surface normals of cut objects. But intersecting before fragmenting does.
• stack_all() fails (possibly memory issues?) for large packings.
• Meshes fail on servers (IBT067 and IBT012) even though the whole toolchain was recompiled. Locally, IBT918, Arch Linux, those same meshes and configs work
  • OCCT 7.5.0-3 (some patches for install location changes in OCCT, but otherwise the same)
  • GMSH 4.8.4 (from git tag gmsh_4_8_4)
  • Fixed (partially) by setting the following in input.yaml::gmsh:
    • Geometry.OCCBoundsUseStl: 1
    • Mesh.StlAngularDeflection: 0.08
    • Mesh.StlLinearDeflection: 0.0005
  • It also seems to be mesh size dependent(???)
• Weird segfaulting with General.NumThreads on IBT012.
  • Crashes at random parts of meshing on some values of General.NumThreads
  • For Delaunay + HXT algos
  • https://nielscautaerts.xyz/debugging-python-in-neovim.html
  • https://stackoverflow.com/questions/10035541/what-causes-a-python-segmentation-fault
  • https://gitlab.onelab.info/gmsh/gmsh/-/issues/1807
  • munmap_chunk(): invalid pointer
  • corrupted double-linked list
  • General.Verbosity = 3,4 makes it segfault
  • python3.9: malloc.c:2542: sysmalloc: Assertion `(old_top == initial_top (av) && old_size == 0) || ((unsigned long) (old_size) >= MINSIZE && prev_inuse (old_top) && ((unsigned long) old_end & (pagesize - 1)) == 0)' failed.

https://gitlab.onelab.info/gmsh/gmsh/-/issues/1246 https://gitlab.onelab.info/gmsh/gmsh/-/issues/1061 https://gitlab.onelab.info/gmsh/gmsh/-/issues/1330 https://gitlab.onelab.info/gmsh/gmsh/-/issues/1480 https://gitlab.onelab.info/gmsh/gmsh/-/issues/338