20

What is the best (scalability and efficiency) algorithms for generating unstructured quad meshes in 2D?

Where can I find a good unstructured quad mesh-generator? (open-source preferred)

Paul
  • 12,045
  • 7
  • 56
  • 129
Allan P. Engsig-Karup
  • 3,226
  • 19
  • 31
  • Tsk. A tip: when creating a multi-word tag, please use dashes to separate out the words. – J. M. Dec 28 '11 at 23:49

2 Answers2

17

The are essentially two approaches to free quad meshing:

  • Direct methods generate a quad mesh directly, usually by some advancing front method. The Paving paper is a standard reference and is the method used by CUBIT, so you have seen these meshes in many publications.

  • Indirect methods generate some intermediate decomposition of the domain (e.g. triangles) and then produce an all-quad mesh through recombination and/or further decomposition. Q-Morph is an example that is used by ANSYS.

Note that smoothing is necessary for both approaches, sometimes with alternating topology fix-up and smoothing steps. Some open source tools have built-in smoothing facilities and the LGPL-licensed Mesquite package is designed as a library specifically for mesh quality improvement.

I know of two open source free-quad meshers:

  • Gmsh (GPL with linking exception) can generate quad meshes using a recombination algorithm described in this paper.
  • The Jaal component of MeshKit (LGPL) is based on recombination similar to Q-Morph above, read the IMR-2011 paper for more details. You can download the source through the link above, but it is not ready for production use yet.
  • LBIE generates quad and hex meshes from volumetric data. From what I can tell, it is an interactive environment rather than a library. The site says that the source is available under GPL upon request.
  • CUBIT is not open source (and although not expensive compared to commercial software, acquiring a license takes a long time), but produces high quality meshes and can be linked into other applications.
Navaro
  • 156
  • 9
Jed Brown
  • 25,650
  • 3
  • 72
  • 130
  • I have learned that Gmsh (http://geuz.org/gmsh/) now also has support for unstructured 2D quad generation by some ”Blossom” recombination algorithm. – Allan P. Engsig-Karup Dec 29 '11 at 09:59
  • Is there a similar method for 3D hexahedrons? – Ondřej Čertík Dec 31 '11 at 22:18
  • 2
    Hex meshes are usually done by sweeping quad meshes and then smoothing (robust, but have to decompose domains). You can also create hex meshes by recombination using H-Morph (Owen and Saigal, 2000; http://www.osti.gov/energycitations/servlets/purl/756062-uQ2PGj/webviewable/756062.pdf), a method similar to Q-Morph referenced above, but it is difficult to maintain quality. – Jed Brown Jan 02 '12 at 00:00
  • 1
    Yes, gmsh has supported quads for a while now. – Wolfgang Bangerth Jan 04 '12 at 04:02
  • This answer is about two years old now and I believe this is still a very pertinent question. Has there been significant evolution on unstructured quad and especially hex meshing? I was wondering if there were any new software (commercial or open source) which allow for 3D unstructured hex meshing. I know gmsh can do 3D hex by extruding 2D quads, but i'm looking for something more appropriate for complex geometries. – BlaB Nov 08 '18 at 16:55
  • @BlaB No revolution in the past seven years. Unstructured hex meshing of complicated geometry is still labor-intensive and/or has quality challenges (see H-Morph). – Jed Brown Nov 08 '18 at 20:01
1
  1. It would be great if gmsh would further include mesquite as 3rd party.
  2. I am from the engineering field, where structured grid is possible and strongly advised.
  3. CUBIT seems to be the mesquite's professional version, or commercial version.
  4. mesquite project ended a few year ago, though some other guys seems to continue keep it going. (meshkit?)
  5. I have a very very good GUI experience with gmsh, it's easy going, light weighted, etc.!
Daniel
  • 133
  • 5