Hi CFX-5 users,

Is it possible, in CFX-Bild, to create "structured" surface mesh and then use AFI tools for volume mesh generation?

Best Regards,

KG

As far as I know it is not possible in an automatic way. In theory you could split up your geometry and create a structured mesh with CFX-4-Build and a unstructured mesh with CFX-5-Build. Then import both and join them using GGI's. This could lead to a lot of tedious work and will probably not help you much because I think the structured nature will be lost when you have imported it in CFX 5.

Astrid

You cannot seed an unstructured mesh from a given surface mesh (not yet at least). Astrid's suggestion is a good one.

Astrid: "I think the structured nature will be lost when you have imported it in CFX 5"

When you import a structured mesh into an unstructured solver, the structured data is thrown away and replaced by a connectivity table. There is really nothing lost here, just a different way to represent the mesh topology.

It is not the structure of the mesh which makes a structured mesh desirable (presumably what you meant by nature). Rather, it is the hexahedral elements used in generating a structured mesh which has some desirable qualities.

Regards, Robin

Or you could use the mesh mode only in CFX-5 and seed as in CFX-4 uqsing Patran Volumeshing to build a mesh of "structured" prisms. This works fine, e.g. if you need to control the number of seeds along your edges. The only drawback is that you have prisms (hexes split in two) instead of hexes. This works very well.

Dear CFX users,

Thank you a lot.

As you can see from my post I put word structured in quotes ("structured"). With "structured" I mean hexahedral surface mesh on my walls. With this approach I can orient my near wall cells in flow direction and with this my numerical error will be minimal. From my experiences convergence will be also faster.

Any new advices (Robin, Astrid, Holiday)?

Best Regards,

KW

Hi KG,
You can use the CFX-BUILD provided by CFX-4. You have a feature in your license file called "CFX-BUILD" and this feature can be used for both CFX-4 and CFX-5 BUILD pre-processors. Try to get a copy from your local cfx support. As far as I know it is not illegal since the feature has the same name for both pre-processors. But I must warn you that the CFX-4 BUILD has some differences regarding creating solids and blocks, but there are a lot of tutorials that might help you in your journey.
Regards, cfd guy

Hi KG,

When you say from your experience, with what code are you referring to?

Robin

Hi Robin,

I'm referring to Fluent and CFX-5 softwares. I have much more expiriencies with Fluent.

BR,

KG

Hi KG,

I think you will find that you get sufficient accuracy with an inflated prismatic layer and run a higher order advection scheme (blend factor of .75 to 1.0, or High Resolution scheme). There is really no need to use a quad mesh at your wall.

I have no experience with Fluent doing this, but it certainly works in CFX-5.

Regards, Robin

Robin,

I agree with your posting. What I meant was that the IJK-directions are lost within CFX 5 because of the connectivity table.

Astrid

Hi Robin,

I wonder how you could easily get the mesh independent solution in unstructured mesh. You have 3 different elements in the unstructured mesh (if compared with structured mesh, you only have to worry about hex elements)....I have a confusion of which elements that I should refine first.......That makes CFX5 a lots difficult to use when come to mesh quality checking (if compared with codes that use structured mesh).........

In this case, once you have what you think is a properly converged solution, you could restart it with mesh adaption on to resolve the criterion that matter to you and ensure that your solution is mesh-independent.

In any case traditional mesh independence methods, such as those advocated by Roache (multiply by two) are okay on structured grids, but I haven't found any method in the literature re unstructured grids. Therefore I think the method I described earlier is the best bet to check this rationally.

Hi Confused,

You are right that it is more difficult to determine the order of grid convergence on an unstructured grid, but the same methodology applies; refine the grid, watever element type, and compare the solutions. When the solution no longer changes, it is no longer grid dependant.

Robin