Survey Results: Mesh Types, Part 2

Check List

The Mesh Types survey included several questions with short-form answers and we shared that data last month. Here we have answers to the long-form, open-ended question about what types of meshes you said you wanted to generate. So let’s start with question 12.

Q12: Share any comments, suggestions or feedback on the types of meshes you generate or want to generate.

“Polyhedral meshing.” The first thing I always ask when people mention polyhedral meshing is whether they’re talking about honeycomb polyhedra or hanging-edge polyhedra. Regardless, if you saw our recent blog article about our intern’s work you saw that he tested our new poly (half-edge, half-face) data structure by creating the Voronoi duals of meshes in Pointwise.


This work-in-progress illustrates our progress toward adding polyhedral support, both honeycomb polys like above and hanging-edge polys, to Pointwise.

This is a work in progress so I’m not certain when this will be released in production code (there’s a lot more work to do) but it (hanging-edge polyhedra) is a logical extension of recent release of hex-core meshing with T-Rex.


Example of hex-core meshing with a T-Rex, boundary layer resolving, near-wall mesh.

“Custom meshing is really helpful.” Yes, indeed it is. I’m hoping you meant having the control to tailor your mesh to your application. Or maybe even writing a template with Glyph scripting to automate your meshing.

“Fast efficient mesh generation with attention to boundary layer embedding.” Have you seen our T-Rex meshing technique that rapidly generates near-wall layers of prisms and/or hexes and then blends to a tet or hex mesh in the farfield?


An illustration of an unstructured mesh with boundary layer embedding generated with Pointwise’s T-Rex technique for a multi-module centrifugal pump.

“If I could have a feature added in my current workflow, it would be to handle going from small to large cells in a very smooth way.” See answer above.

“Prism layer generation is the greatest challenge in most cases.” See answer above. Also, your surface mesh is the key to a good volume mesh.


“Quad dominated surface, unstructured.” Got that. And structured quads too.

1024x614_Webinar - Unstructured Quad & Hex – 2016

Example of a quad-dominant surface mesh from which hex layers are extruded using T-Rex.

1028x618 _Webinar - Multi-block S-Duct - 2017

Example of a multi-block structured grid for a diffusing S-duct.

“Mesh adaptive refinement.” Mesh adaptation, including surface mesh adaptation that retains adherence to the underlying geometry model, is something we’ve been working on for quite some time. Our approach is based on use of a sparse dataset from the flow solver (a point cloud) that we use as a clustering source.


Initial (left) and adapted (right) mesh for the upper surface of the ONERA M-6 wing. This image is from a case study demonstrating Pointwise’s approach to mesh adaptation.

“Blocked structured mesh is very difficult to generate for complex geometry.” Back in the 80s it was all we had and we liked it!

“Why aren’t FEA meshes included in this pole [sic]?” Because we serve primarily the CFD world. But after rereading the questions someone doing FEA (by which I assume you mean that your application is solid mechanics) could’ve answered them.

“Hexahedral non-conforming meshes with hanging nodes for localized mesh refinement.” See answer above about polyhedra support.

“Polyhedral mesh with layers.” See answer above. Also, which kind of polys?

“Higher Order Meshes.” This has been an active area of research at Pointwise for several years and in 2018 we launched the first, commercially available, high-order mesher for CFD. You can read our hype here. But don’t take our word for it. You can watch a webinar from a customer about our H-O meshing in action.

“Combination of unstructured quads and structured mesh for 2D.” Yep, we’ve got all three: unstructured, structured, and 2D. Pointwise’s mesh topology lets you apply different meshing techniques in different regions. So one surface mesh can be structured quads, its neighbor can be unstructured quads, its neighbor could be triangles, and its neighbor could be quad dominant.

“Please think about easy to read and write neutral format that can be used to share meshes between different solvers.” Easy is a relative term but you should consider CGNS which was designed for precisely for the purpose of portability. Otherwise, I can’t think of a single format which supports all CFD solvers. And that is the reason why we created a Plugin SDK for reading and writing mesh data so you can write your own reader and/or writer for precisely the format you want (and make that format as easy as you want).

Check out our Ultimate Guide to Writing Import and Export Plugins.

“Survey only allows for one CFD code technology, most people use more than 1. Also did not represent SPH solvers.” You are correct. As I mentioned in Part 1, writing a good survey is difficult. With time I hope to get better.

“Interested in generating high order meshes for finite element solvers.” See answer above.

“Hanging node with the voxel meshing would be great.” I agree 100% which is why we’re working on it. See above.

“Polyhedral meshing would be nice.” It sure would be. Which is why we’re working on it. See above

And finally…

“I want to be free from Meshing bondages.” I have that dream too. Free at last, free at last, great God almighty, free at last.*

ICYMI, here’s Part 1 of our Mesh Types Survey results which includes all the numbers and charts.

Let Pointwise free you from your meshing bondage. Request a free trial today.




*With all due respect to Dr. Martin Luther King Jr. and I Have a Dream, one of the greatest speeches of all time.

This entry was posted in Applications, Software and tagged , , , , , , , , , , . Bookmark the permalink.

4 Responses to Survey Results: Mesh Types, Part 2

  1. Jeremiah says:

    I have two comments, which I don’t think I included when I completed the survey. It would be nice to control boundary decay and T-Rex parameters for individual boundaries of a domain/block. In other words, maybe I want a growth rate of 1.3 from one surface, and constant height (GR = 1.0) on another surface.

    Second, would it be possible to support T-Rex growth rate < 1.0? There have been a few cases where I would have liked that. I suppose there are other issues that would arise, however.

    • John Chawner says:

      Thanks, Jeremiah. Take some comfort in knowing that you’re not alone in requesting those things. I’ll update our requirements database with your name.

  2. Pingback: This Week in CFD | Another Fine Mesh

  3. Pingback: Survey Results: Mesh Types, Part 1 | Another Fine Mesh

Leave a Reply