[jbritton]

Hi All,

Quick question mainly out of curiousity.

In Turbo machinary the trend is to create nice pretty structured hex grids in order to carry out analyses. But my question is that while I can see this being improtant for a code such as FLUENT Im unsure as to the advantages for a FEM based solver like CFX.

The solver in CFX is based on the central node of the elements rather than the shape of the cell and Im unsure as to why you would use a structured hex grid as opposed to an unstructured mesh, and what the differences in accuracy are?

Thanks

[noslip]

the cfx solver uses the nodes to create control volumes around it. i assume by "unstructured" you mean tetra-meshes (ansys uses an unstructured approach even for hexas)?

the big advantage of using hexa meshes in turbomachinery applications is that you can align your mesh relatively nicely with the flow direction, therefore reducing numerical diffusion and aiding convergence and you need less elements to fill your domain at the same time (this applies not just to cfx).

in situations where the flow is more along the lines of chaotic rather than guided, hexa vs. tetra doesn't really matter anymore (apart from the amount of elements needed).

[murx]

I want to come back to this question.
What is the disadvantage of automatically generated hexa-dominant meshes compared to hexahedral meshes built from a block-structure when using an unstructured solver?

Right now I use blockstructured hexa meshes in CFX. The quality of some cells is not very good, plus local mesh refinement is limited due to the propagation of the refinement over the neighbouring blocks. I wonder if I could benefit from those hexa-dominant meshes since I use an unstructured solver anyway.

[ghorrocks]

The answer to that depends on the application. You really have to try it on your application and see if it helps. If hex dominant meshes were better than structured meshes all the time then nobody would bother with structured meshes.