I would like to know the latest available Gambit version. Thank you in advance, Thomas Farmakis, MD, Cardiologist.

As far I know the latest version is the 1.2.4 released an March. In fluent's site you could find more details.

Correct, 1.2.4 is the latest release. Gambit 1.3 will be released in May, with new STEP and Parasolid import and export, plus a few other enhancements.

Scott,

I was wondering what Fluent's plans are regarding automatic hexahedral grid generation for complex geometries? I realise that Delaunay triangulation is commonly used to generate unstructured meshes with triangular (tetrahedral in 3-D). But what sort of techniques are used to perform automatic hexahedral meshing (aside from paving, if any)?

Thanks for your time.

Glenn,

We're working on two different technologies for automatic hex meshing in complex geometries. One is an extension of Gambit's Cooper tool to handle multiple sweep axes. This will generalize the Cooper tool to make it applicable to a much broader range of geometries. The other is our "GOCARTS" technology, which creates a mesh which is Cartesian in the bulk flow region, with hanging-node refinement toward the surfaces, on which boundary-fitted layers are automatically created. The two methods are complementary, each being more beneficial in certain types of applications. We are also following research progress with other techniques, including Whisker Weaving, Plastering, and medial-axis methods. Several codes are using these latter methods with mixed success.

Scott

Hi Scott

What I would really like to see in Gambit is more flexibility with Hex meshing. For example the ability to move nodes, not just on faces but in volumes and to be able to fit boundary conditions to individual cell faces. I also find that having a boundary layer that can only have integer multipliers of cells are very restrictive, e.g. 10 cells cannot be reduced to 5. This in my opinion is a major drawback. What about incorporating integral and arbitrary coupling of cells this would make meshing far easier, and better.

I find Gambit very time consuming building meshes as in order to build a mesh one has to draw a geometrical line select it at least another two times. Why can't we have a 'face of cells' that can be extruded and rotated at the same time giving a volume of cells.

The ability to move vertices around after they have geometry and cells attached to them would probably save a lot of time without having to redo everything. Journal files can be a pain.

The 'ultimate wish' would be the ability to manipulate geometry after it has been read into Fluent.

The technology exists (for years), is there any possibility of any of these wishes could be incorporated.

Allan

Hy allan,

I agree with you in all points. A good meshing-tool would be a combination of Gambit and I-DEAS. In I-DAES you have the possibility of manipulating each node, you can extrude edges or faces of existing cells...and you can create cells by hand and copy, rotate, move, mirror... them. The disadvantage of I-DEAS is the bad quality of the cells with automatic-meshing. Boundary layers: The simplest way would be (example is 2d): 1. and 2. row rectangles, 3. row alternating triangles and rectangles (->half of the cells), 4. row alternating triangles and rectangles (->half of the cells ->1/4 of boundary-cells).

In the last months I also saw sometimes the question: Why to make a perfect 3-d-cad-model with tolerances of 0.000000..mm if I mesh it with cells of 0.1 or 1 ? Isn't there perhaps another philosophy ? Is the "solid-model"-philosophy really good ? No question: it is state-of-the-art in cad, but is it really also the best and only for meshing ? Let's make innovations ...

Jurek

I thought we had moved on from manual mesh generation methods to automatic mesh generation - to relieve the use from the dreaded "Mesh Monkey Finger" ailment.

Maybe you would like to try your advanced 2 Dimensional mesh generation technology to a relatively simple 3 Dimensional solid model to see if it works - I would suggest that it won't. Unless you are looking at 2 Dimensional duct flows that is!

Solid models exist to aid the design process and engineer in a concurrent engineering environment, not just so the analyst can create a MESH. Analysis models are simplifications of the real things therefore the tolerance at which the solid model is created is not an issue.

Techniques exist within solid modelling packages such as I-DEAS to suppress design features considered to detailed for the analysis process. Suppression of features to aid meshing is a trivial process. Re-inventing the wheel (creating another model) to aid mesh generation is a poor approach to mesh generation.

Wake up to Real Worl Analysis!!!!!

P.S. has anyone got any cream for "mesh monkey finger"

Automatic mesh generation is fine but I think it's still very useful to be able to manipulate certain vertices and cells to be able to make small/medium geometrical changes, which is often the case in development work, without having to completely redefine the whole case, which wastes time and money.

There are other cfd programs which have the ability to do this and it is my experience of them which allows me to make these comments.

I am not refering to simple 2-d shapes here but highly unstructured hexahedral meshes in 3-d geometries.

allan

Maybe you would like to try your advanced 2 Dimensional mesh generation technology to a relatively simple 3 Dimensional solid model to see if it works - I would suggest that it won't. Unless you are looking at 2 Dimensional duct flows that is!

It would work in every 2-dim geometry: 2 rows of rect. then 2x rect+trian, then trian or rect (pave) for the rest.

3-dim: if you extrude the 2-dim object from above, you have (very simple) 3-d-object. What I want to say: Even in complicated 3-d-objects you could divide up the geometry in "meshable quasi-2-d-parts ". Another idea: 3-d-geometry, some layers of structured hex, then the wedges/hexes. Now it is "streched and reduced" in one direction. Now the same in the other direction. And the volume and "unmeshabale regions" with pyramids and tets.

Solid models exist to aid the design process and engineer in a concurrent engineering environment, not just so the analyst can create a MESH. Analysis models are simplifications of the real things therefore the tolerance at which the solid model is created is not an issue.

Techniques exist within solid modelling packages such as I-DEAS to suppress design features considered to detailed for the analysis process. Suppression of features to aid meshing is a trivial process. This works well, if you have easy constructions and you have a really "full" construction". And certainly: suppression of a small hole is not the problem. But what is, if you have several parts ? And the import in gambit does not work ?

Re-inventing the wheel (creating another model) to aid mesh generation is a poor approach to mesh generation.

I think, this is pure theory !!!! Sometimes it works, sometimes it does not. And for this case we need more effectiv methods for creating the geometry. It must not be a conservative cad-tool but something adapted for meshing-purposes.

Wake up to Real Worl Analysis!!!!!

Are you from SDRC ?

P.S. has anyone got any cream for "mesh monkey finger"

For relaxation-purposes in ideas: create a sphere with r=1000 and mesh with tets (third order) with 0.1 mm. Press "mesh" and relax.

At present we are using 2.0.4 version of Gambit.