CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   hybrid mesh (http://www.cfd-online.com/Forums/fluent/41936-hybrid-mesh.html)

Fabrizio August 4, 2006 11:36

hybrid mesh
 
Hello guys,

I'd like to create an hybrid mesh with hexa and tetra.

I created my structured hexa mesh in gambit, my problem is that I don't know how to keep the same surface mesh at the interface between the structured mesh and the tetra mesh. (I want the tetra mesh to wrap the structured mesh and so the interface is more than one).

Every interface consists of a set of surfaces at the structured mesh side which are to be coupled with one surface at the tetra mesh side.

In the past I used to do similar tasks by using "split volume" which allows to create two volume with some faces in common that ensure the correct coupling, but in my geometry I have some virtual entities and it is not allowed to use this technique.

I guess that there must be a simple way to do this.

Thank you very much for any hint!


Jason August 4, 2006 13:09

Re: hybrid mesh
 
You've got two relatively simple options (as I see it... I might be missing other options).

The first is to manually connect the paired faces. In the face commands there is a button that looks like a black plug. You then pick the two faces that are to be shared. If the geometry's virtual then you'll have to use the virtual options. This only works if the faces to be connected have the same number of nodes and edges, and these nodes and edges match within tolerance (if you choose the normal virtual option, it uses the default tolerance of 1e-5, I believe, otherwise there's an option to use virtual geometry and specify the tolerance but there may be some face warpage if you use too large of a tolerance).

The other option is to use non-conformal meshing. For this one you simply mesh all of the geometry, and don't worry about the mesh being linked across the structured and unstructured faces. You do have to be careful that there isn't a large mesh size difference across the boundary though. When you're defining the BCs, the faces that should've been shared but belong to the structured mesh are given one "interface" BC, and the faces that should've been shared but belong to the unstructured mesh are given another "interface" BC. Then read the mesh into Fluent and define the non-conformal interface (Define->Grid Interfaces... don't use the periodic or the coupled options).

The grid interface tends to be easier, but you're going to incorporate some numerical error... if your grid is refined enough and there isn't a large change in mesh size across the interface, then the numerical error will be small or negligible. Connecting the faces is usually the "correct" way of doing it, but if you have a lot of faces, or if they have different numbers of nodes (or the nodes don't line up...) then this can become very tedious, very quick, or even impossible when dealing with virtual geometry.

Hope this helps, and good luck, Jason

Fabrizio August 7, 2006 02:31

Re: hybrid mesh
 
Thank you very much, Jason.

Now everything is more clear. Even if this is a little disappointing: it doesn't exist a way to do everything both quick and clean...

Fabrizio


san August 9, 2006 01:10

Re: hybrid mesh
 
i am intrested to know more amout hybrid mesh. It will very appriciable,If you can share any tutorial which explains hybrid mesh.

thanks inadvance san


kenischi August 16, 2006 06:41

Re: hybrid mesh
 
*********************************************

just an easy way:

- mesh one volumne structured

- mesh second volumne unstructured

- copy second to somewhere with mesh

- delete basic second unstructured

- recopy copied to old place

- you have two faces for interfacing, also ideal cell sizes for structured and unstructured

- but better way is setting tetrahedral cells between str.-unstr. in tgrid

As i know interfacing is just a 1-d surface interpolation, which destroys vector directions of the 2 missing directions

regards optixxx


All times are GMT -4. The time now is 03:02.