CFD Online Discussion Forums

CFD Online Discussion Forums (
-   FLUENT (
-   -   Beginner (

T.Rajkumar June 28, 2006 06:58

Hello friends

iam a beginner in CFD and Fluent. i have certain doubts.please help me.

1. I came to know that Hexahedral meshes give good result (accurate)than tetrahedral meshes.How this is justified. 2. what is physical importance of size functions in GAMBIT.

Amr June 28, 2006 12:12

Re: Beginner
Hi Raj,

As for your first question, it is all about discretization errors that happen when fluent solves your equations. Discretization means that Fluent is approximating the change in variables accross each cell in a 1st, 2nd or 3rd order fashion depending on your solver settings. Now, when it comes to cell structure, hex cells do better in reducing the truncation error -that results from approximations across cell face or volume- and furnishing smoother "change" in the variable you solve. This is primarily due to their geometrical features especially when with low skewness values.

Unlike hex meshes, tet cells do not possess the geometrical features that allow them to have this smoother change in your variables. No matter why equilateral triangles (or prisms in 3D) are best to use in case you fail in having or your cells as hexahedras. Numerical diffusion is the term that describes the resulting truncation error from above. If you refer to Fluent manual, in the grid section, you will see that numerical diffusion is mitigated by using hex meshes in general.

Now, does a tet grid provide less accurate solution than a hex one? The answer relies largly on your flow physics. In regions where variables do change steeply accross a cell volume/face, it would be better if you choose your cells so that you guarantee this smooth change either by utilizing hex low skewed cells or equilateral -low skewed- tet cells.

Having said that, there is no point in that tet cells always provide inaccurate solutions since numerical diffusion depends on your flow features and associated gradiests. Since they are robust, easy to build and suit vertually all complex geometries, tet cells are oftentimes your only way to construct your grid. Code developers are working hard to introduce new numerical techniques that would overcome numerical diffusion, thus, allowing commercial users to easily construct their grids while not fearing from inaccuracy.

This was to the best of my knowledge. Hope this is correct and helps!

Cheers, Amr

T.Rajkumar June 29, 2006 09:40

Re: Beginner
Thank you Very much Mr. Amr

Iam satisfied with the answer.Please answer for the following query also.

1. Physical significance of giving size functions. 2. what is Volume Decomposition.why it is done.

Thank you. Raj.

Amr July 3, 2006 20:48

Re: Beginner
Hi Raj,

I would recommend that you check the gambit user guide in regards to your questions. As for the decomposition, volume decomposition is utilized in gambit so that you may have the optimum grid structure you look for. As you may know, gambit uses certain mathematical algorithms that produce the mesh automatically inside your geometry like map, submap, pave, cooper, T-grid schemes. Those shemes depend on certain geometric features and vertex types as detailed in the gambit user guide.

Map scheme, for example, produces hex meshes. So, if you wished that all you geometry be meshed with that map scheme, you should make sure your geometry is map scheme compatible. One important way of doing this is by decomposing your geometry to "mapable parts" (i.e. partitioning your geometry) so that gambit could produce hex meshes with the map scheme. This philosophy applies to all other meshing schemes.

For the size functions, yes the do have physical importance. They are used to vary the cell "grid" sizes within your domain. It enables you to build coarser grid where -in your domain- you expect weak changes in solution variables, and on other hand, build finer grid where you need to resolve strong changes in your solution variables.

All the above is very well discussed in the gambit user guide. I hope this helps!

Cheers, Amr

All times are GMT -4. The time now is 01:46.