|
[Sponsors] |
August 7, 2003, 11:46 |
Ungent
|
#1 |
Guest
Posts: n/a
|
Dear All
I am simulating the flow through a tube, in which I used the parallel mesh style for border and a triagnle style for lumen in Gambit? Who does know the advantages of the parallel mesh style in comparison with the triangle mesh style for the tube border. How should I select the mesh style for the border in a tube. Thank you very much. |
|
August 8, 2003, 13:58 |
Re: Ungent
|
#2 |
Guest
Posts: n/a
|
I think you refers to structured hexaedral grids when you talk about parallel mesh, right?
Structured grids allows you to reduce numerical diffusion, if cells are aligned with the flow, which is impossible using unstructured grids. Also, if your geometry is simple, a structured grid usually has less nodes, so it allows you to reduce you computational time. If your geometry is complex, however, genereting a structured mesh is difficult and time consuming, because you have to find the proper domain decomposition, which allows you to apply a hex mesh to all its parts. If you have to model a single pipe, you could try using a structured hexaedral mesh. Many books suggest to split the volume in a central brick and in four lateral parts, in order to have a uniform grid in the center of the pipe. The section will be something like this (sorry for poor quality, hope you understand) \_/ (|_|) / \ Mesh the internal edges, then apply the submap scheme to the lateral parts of the section and the map scheme to the central square. Specify how much nodes you want along the axis of the pipe, then simply mesh the five volumes in which your cilinder is splitted. Hope this will be useful Hi ap |
|
August 8, 2003, 18:51 |
Re: Ungent
|
#3 |
Guest
Posts: n/a
|
Thank you very much. I am going to mesh a branched tube for fluid flow simulation inside tube. Each branch has less diameter than the trunk. Do I need to split it into three parts from the branching point?
|
|
August 8, 2003, 23:33 |
Re: Ungent
|
#4 |
Guest
Posts: n/a
|
Meshing a branched pipe using structured grids is not easy.
If you have a T intersection, you can find an example in GAMBIT tutorials. There should be more info on some CFD book. If you have something like >-- it's easier and faster to use an unstructured mesh, which allows to mesh your domain almost directly. You can specify the number of nodes you want along its edges. Hi ap |
|
August 9, 2003, 04:31 |
Re: Ungent
|
#5 |
Guest
Posts: n/a
|
I have experience that solution on structured grid is easier to attain.
Cathy |
|
August 11, 2003, 06:48 |
Re: Ungent
|
#6 |
Guest
Posts: n/a
|
Sometimes unstructured grids have some convergence problem.
However, if the quality of the mesh is good enough, you shouldn't notice particular difficulties obtaining the solution. Hi ap |
|
|
|