Hi there, happy holidays!

On the market of CFD codes, FLUENT, CFX, Star-CD, etc.sale quite well. Now, let's forget their solvers, just make a comparison on the functionaliy of their pre-processors, which code is with the best pre-processor? Comments from whoever has experience with at least two of those commercial codes are very appreciated!

Hello all, Gambit is the best compared to build. bye

According to my experience GAMBIT is the best.

Which is best depends very much on what you want to do:

What kind of geometry source do you have (CAD?)?

Do you want to do unstructured or structured meshes?

Do you want to do resolved boundary layers?

How large meshes do you want to be able to create?

Will you do a lot of repetitive meshing (changing a small parameter and re-meshing a similar geometry)?

Jonas, Let's start without any geometry source, both structured and unstructured mesh, the mesh should be fine with or without concern on boundary layer, and mesh size should be proper to industrial applications, such as half a million or one million, then which pre-processor is the best?

y r u only concerned abou the pre-processor. Y can't u consider the solver part,where fluent losses it's point.

The distinction in classification between structured and unstructured meshes has always seemed a bit daft to me. I realise that there are obvious solver related issues but many people, especially when evaluating CFD tools, bother too much about whether the grid is structured on unstrcutured without appreciating how the differences in these 2 types may affect their analysis. In fact, I'm not at all sure that the differences in these gridding system will affect the analysis at all in the way they imagine.

Surely it is better to ask question such as whether the geometry under consideration is highly curved or cluttered or cartesian etc. Then maybe a choice should be made of the appropriate gridding system...??

Could it be a question of mesh quality or viscous boundary layer resolution?

Maybe I'm missing something? Just how important is it that a cell's neighbours can be located by simple i+1, i-1 type methods or not? How can this affect the choice of structured or unstructured apart from the dependency of the available solver?

Fred.

Gambit starts off easier and has nice functions to make (relatively) easy geometries. The problem comes when you 'real-life' geometries from CAD. The clean-up tools in Gambit are non-trivial and dont work very consistantly.

CFX-Build5.5 is much more for CAD-literate people and reads in IGES files very well. Also, there are strong clean-up tools.

More importantly though, is the mesh generation. I think it is easy to get a mesh from any pre-processor, the difficulty is getting a good mesh. The tools for refinement are poor in Gambit (i.e. using sub-domains) whilst they are strong in CFXBuild5 (using mesh controls).

Lastly, but most importantly of all is the adaption. I think we should all using mesh adaption as a matter of course.

Gambits adaption is cumbersome and very manual whilst CFXbuild5.5 has automatic adaption which you only need to define which variable to adapt to and how much you wish to adapt.

In summary:

Gambit: Looks great, easy to use, but stick to geometric representations and not CAD.

CFXBuild5.5: Less swish, but very strong for CAD and meshing.

Choose at your will, but remember, the simple things that you might do on a trial are not what your are going to be when you have an engineering problem to solve.