CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   Hybrid mesh generation (

Jake April 16, 2007 14:19

Hybrid mesh generation
I have been tasked with moderating a session on hybrid grid generation techniques and need to come up with some talking points. As I am not very familiar with the hybrid gridding strategy I am hoping to get some input from this community. I am most interested in answers to the following questions:

How long does it take to generate an acceptable hybrid grid compared to other grid strategies?

How long does it take to converge a CFD solution on a hybrid grid compared to other strategies. For this question, the grids used for comparison should provide the same "accuracy" of CFD solution.

Which commercial or open packages are good at hybrid grid generation? Which is best?

Are there any situations where hybrid grids can't be used?

Are there any other issues I am overlooking?

By "other grid strategies" I include overset (structured and unstructured) and non-overlapping block structured.

Thanks in advance for your input.


Pagey April 16, 2007 18:12

Re: Hybrid mesh generation
Talk to the people at ARA Ltd., they do a lot of hybrid meshing.

John Chawner April 21, 2007 14:27

Re: Hybrid mesh generation

Your hybrid meshing session should be pretty interesting given the current state of the art. For the record, I'll assume hybrid means prism-tet or hex-pyramid-tet, or hex-prism-pyramid-tet. Structured means mapped hexes. Unstructured means tets. Furthermore, I'm going to assume that we're talking about viscous as opposed to inviscid simulations.

> How long does it take to generate an acceptable hybrid grid
: compared to other grid strategies?

This question might be easy to answer if not for the word "acceptable". In general, I would say that generating a hybrid mesh is faster than generating a structured grid but slower than generating an unstructured mesh.

If acceptable implies accuracy, I know people who refuse to use hybrid meshes on the grounds that they produce inaccurate results. If acceptable implies efficiency, I know people who refuse to use hybrid meshes because the solvers are slower and use more memory. In both cases, these people use structured grids instead. And I'm not talking about single users; these are entire companies.

Take a look at the results from the most recent AIAA Drag Prediction Workshop. That should give you some background on the various meshing strategies and solution techniques.

The promise of hybrid meshes relative to structured grids is the ability to generate a mesh for a viscous simulation much more rapidly. The price one pays for that level of automation is having to deal with quirks in the resulting mesh.

All times are GMT -4. The time now is 11:46.