Hi,

I was trying Tetra meshing in CFX-Mesh and ICEM. I noticed that 'Advancing Front'; the meshing algorithm employed in CFX-Mesh gives a mesh quality of this type

http://img2.freeimagehosting.net/ima...dcd605f2eb.png

While the Octree technique in ICEM gives a mesh quality of http://img2.freeimagehosting.net/ima...47e6bc4e5e.png

It looks to me that Octree gives more or less structured mesh in interior volume regions. Is it? It would be great if some one can share me with his/her experience with these two type of mesh techniques. Merits and demerits.

Many Thanks in advance, Sam.

Hi,

My experience with Advancing Front is that it requires less user input to generate a good mesh, but it does require good quality CAD surfaces and tricky surfaces sometimes don't mesh.

Octree can require more user input for smoothing to a good quality mesh, but in ICEM it can mesh very poor CAD and tricky surfaces.

Also the Octree approach generates a very regular internal mesh which converts to hexas nicely if you want to generate a hexa core mesh. Advancing Front meshes are not as good for this. The internal Octree mesh is not structured, it is still unstructured just with a regualr repeating pattern.

Glenn Horrocks

Thank you very much Glenn. You gave me a very good insight of both the techniques.

I request you to please clarify a few more doubts of mine.

1) I guess CFX is an unstructured solver. Does structured mesh will still have some benefit over unstructured mesh? Structured mesh takes more time to be generated so I want to know if it is worth investing time in creating the same.

2) From various threads, I see that number of elements in a mesh affects the number of integration points and hence the equation assembly. So solver speed will also depend upon it. Tetra elements have more elements per node than Hexa elements. So tetra will take more time to to run the simulation, right? But I am wondering why HEX mesh is expected to give better results in general while it has less integration points than TETRA mesh. As a beginner, I might think that higher the number of integration points, better the accuracy of results. But that's not quite seen. Why?

3) Fluent manual suggests to prefer Hexa mesh for LES simulations. While I got the impression from CFX manual (First paragraph Page 113 Modeling Guide V11) that TETRA mesh will be better for LES simulations because hex elements might suffer from being anisotropic. Please help me in deciding the type of mesh to prefer.

Sorry for asking too many questions, your replies always help up to a great extent !

Many Thanks & Regards, Sam

Hi,

Question 1) CFX is an unstructured solver. A structured mesh can still be a benefit to CFX as a structured mesh can be of higher quality than an unstructured mesh.

2) A tet mesh will result in a larger memory requirement (typically double) and slower run time (typically 50% more) than a hexa mesh with equivalent edge lengths. This assumes both meshes are good quality elements.

3) LES is especially sensitive to mesh quality. In particular it suffers from numerical dissipation from high aspect ratio or non-orthogonal elements. LES is much happier on hexa elements with an aspect ratio of 1 and orthogonal faces.

Glenn Horrocks

Regarding Octree, I think that it should be used to create a preliminary mesh. THen, smooth the resulting surface mesh, and then, use advancing front or Delaunay to create the volume mesh. Using Octree makes tetras with a volume ratio of 2, while with advancing front or delaunay you can choose the growing ratio. Bye.