Hi, everybody,

I have worked in CFD with structured grid for many years. Recently, I have to learn to use unstructured grid because my job's reason. I want to have a whole picture about the method. however, I am quite confused about the unstructured grid. It seems that bunch of manners and new nomenclatures, such as: tetrahedron, hexagon, and finite element method etc.,are involved in the techniques. Could you please give me some advises, how to start to learn it. Is there some websites helpful?

thanks Ben --

Hello,

Unstructued mesh clearly has an advantage over structured when we make mesh. It's much faster even for complicated geometry. The main difference in the mesh data of the unstructured mesh is it has connectivity among elements in addition to coordinates.

But it usually makes matrix worse than structured does. In some cases, such as highly nonlinear physics, biomedical analysis or crash simulation of metal, we should use structured. It's inevitable.

I prefer structured mesh because it pays off after all when the accuracy of simulation matters. I might be against the main stream. I easily spend 4-8 weeeks to make one good mesh, i.e. cylinder block with lots of holes through the walls. It seems to me, more and more people return to structured mesh because there are mesh generation tools which can handle extremely complicated geometry and thousands of surface data.

I do not understand the motive of your interest in unstructured mesh. No job for structured?

To find the information, you don't have to go far. There are lots of discussions in this seemingly unstructured forum. You could get structured information depending on your effort from this forum. Just type, "unstructured" in the search box.

Selina Tracy

Wow, it takes only 8 weeks! I cost me around 1/2 year. Are you using a commercial or specialized in-house code?

-Amit Bhutani

I usually use a commecial for complex geometry though I've been trying to make my mesher much better. -Selina

Ben,

These two sites have good material on grid generation:

http://www-users.informatik.rwth-aac...eneration.html

http://www.andrew.cmu.edu/user/sowen/mesh.html

Jonathan Shewchuk's site has good information (notes & software with source code)

http://www.cs.berkeley.edu/~jrs/

Regards.

Thanks for your respondes. I have worked in numerical simulation of rocket motors, from chamber through nozzle. It is a kind of the complicated situations for CFD. For some simple configuration within the chamber, we can use structued grid as I have used. However, for some real cases, the geometery is quite complex, I saw many researchers, typically in NASA, etc. are using unstructured. We want to open another new way for our research. By the ways, we usually used some codes written by`ourselves. Sometimes we did use some commerical one for comparision with our codes.

Ben,

You may want to consider looking at the best of both worlds. A popular commercial code that can do both unstructured and structured grid gen is icem cfd, http://www.icemcfd.com

I think unstructured mesh generation is more time consuming, less controllable and harder to post process. Any small change in geometry can easily lead to a different mesh which makes good comparisions difficult. With a structured mesh you can ensure more consistent meshes. Finally the solution is typically less accurate because all the discretization schemes that I know of are much more accurate when the mesh is aligned with the flow: impossible with unstructured. Some geometries are just very unsuited to unstructured mesh: bascially anything with a pipe in it is probably unsuitable for unstructured meshing.

Unless there are just WAY too many small features to resolve, I will avoid unstructured. Typically if there's too many features to resolve with a structured mesh, you will need more mesh points than you can run in a realistic time anyway, so it's time to downsize the domain you model. Maybe in 10 years when machines are that much faster, the problem can have enough mesh points that it would take too long to mesh with a structured grid.

Hi, everybody, I am student and doing MS (POST GRADUATION) in CFD. I am interested to do my thesis work / research in UN -structured grid.Since every thing is new for me, therefore I am confused and not been able to decide area of work (of course in un-structurted grid).can any body help me to pick / select any area for my thesis from un-structured grid (where I can work initially as MS student and then same can be extended to PHD). Thans.

In fluent you have the volume-elements: tets (4 nodes, 4 faces), wedges (extruded triangle: 6 nodes, 2 triangle-faces, 3 quad-faces), hexaeder (as you know..), pyramids (rectangular base and 4 triangles).

With tets you could mesh allmost every solid-body by push-button. Advantage: fast, disadvantage: bad elements

finite element method has nothing to do with the meshing..

If you know structured meshing, it is really no problem to learn unstructured meshing.