CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   ANSYS Meshing & Geometry (http://www.cfd-online.com/Forums/ansys-meshing/)
-   -   [ICEM] Meshing problem in Formula one (http://www.cfd-online.com/Forums/ansys-meshing/110911-meshing-problem-formula-one.html)

Far December 25, 2012 06:19

Meshing problem in Formula one
 
I am working on meshing of formula one with simplified geometry. Geometry consists of:

1. Tyre

2. Three wings

3. Four zero thickness plates.

I am successful in meshing the domain with octree and then deleted volume meshing and proceed with denauly (Tglid and AF). But I have problem when I turn on the prism option in volume meshing.

What do you suggest? Should I use Tgrid for volume meshing for good quality of prisms?

Far January 12, 2013 00:55

Now I am able to get the required meshing through patch dependent method but still facing few problems. I will describe them later in next post. But my question is that why I have to set many parameters to get this mesh, which on the other hand is very simple to make in Gambit!!! :o


http://imageshack.us/a/img59/5020/shellmeshing.png


I have few questions:

1. How to get working the octree method to get this sort of meshing on the surfaces? Which parameters need to be set and I am really confused in this area?

2. Why there is difference in patch independent and patch dependent shell meshing? May I know the backgroud algorithm?

3. I also think that user/help material should be improved. Something like Ansys meshing or gambit help material :)

diamondx January 12, 2013 01:14

a lot of people tend to delete the volume mesh and replace it with delaunay. while generating a delaunay on the octree mesh is a better approach as it give better results. this is an interesting thread anyway. may you share the geometry see what you did to it ? that picture is not enough for me... :D
I'm about to have some rest, i'm coming back here tomorrow morning

Far January 12, 2013 02:38

Ali

I also tend to use this method for volume meshing.

But I am wondering about the surface mesh (aka shell meshing). As you know ICEM gives you two methods to make the surface mesh:

1. Octree method that is top down method. Delete the volume meshing. Smooth surface mesh and fill the volume using the Denauly method. But here two problems arises:

a) surface mesh has problems and even using the edit mesh tools, I am not able to rectify them.

b) Need to set the complicated set of parameters. And I dont understand them. Hope to get them one day :confused:

2. Surface meshing and denauly meshing. That is bottom up approach. Here ICEM provides two options to get the surface mesh

a) Patch dependent
This is working now. But always get the problem in innder region while the mesh at the boundary is fixed through curve mesh parameters. I am also using the following options:

i) respect line elements (is this necessary?)

ii) protect boundary elements

iii) project to surface

iv) adpat mesh interior (affected by surface mesh parameters and succesfully used it in this model. Got idea from the one of the Simon's post)

v) Try harder 1 and some times 3. But dont see any effect in current method

b) Patch indepdent
This method prodecues the different mesh, mesh density and have problems on the surface. It is simply now following surface. Why this method is produceing the differnt mesh than the patch dependent if one patch is being meshed at a time:confused:.

http://imageshack.us/a/img832/9650/shellmeshing2.png

http://imageshack.us/a/img152/4766/shellmeshing3.png

http://imageshack.us/a/img819/4711/shellmeshing4.png

cfd seeker January 12, 2013 10:07

Octree method uses Patch Independent method to project the volume mesh on the surface. Generate the surface mesh first using Patch Dependent method and then filling the volume with Delaunay if you dont want to use PI surface mesh

Far January 12, 2013 10:11

but how the patch independent and patach dependent shell meshing work? I am experimenting this on the single patch (one surace) so the mesh should be same? Right? But it is not.

Another side is that if I make the octree mesh, will I get the same surface mesh on that patch, which I am getting through patch indepent shell meshing?

cfd seeker January 12, 2013 11:40

Quote:

but how the patch independent and patach dependent shell meshing work? I am experimenting this on the single patch (one surace) so the mesh should be same? Right? But it is not.
I do not know the algo behind the two methods but you cannot get the same surface mesh with both. Patch independent dont respect the nose distribution on the curves while Patch dependent does respect

Quote:

Another side is that if I make the octree mesh, will I get the same surface mesh on that patch, which I am getting through patch indepent shell meshing?
I guess so

Far January 12, 2013 12:23

Quote:

Originally Posted by cfd seeker (Post 401499)
I do not know the algo behind the two methods but you cannot get the same surface mesh with both. Patch independent dont respect the nose distribution on the curves while Patch dependent does respect


I guess so

PI also respects the node distritubatin on the curves, if it is asked to do so :rolleyes:

cfd seeker January 12, 2013 12:30

Quote:

PI also respects the node distritubatin on the curves, if it is asked to do so
I am confusing blw the two but one respect and one does-not.....

Far January 12, 2013 12:58

Quote:

Originally Posted by cfd seeker (Post 401506)
I am confusing blw the two but one respect and one does-not.....

Patch dependent is designed to mesh using the curve info. While PI by default does not so. But if we turnn on the option "respect line elements" then it does so.

diamondx January 12, 2013 14:55

https://www.sharcnet.ca/Software/Flu...algorithm.html
can you provide the tin ...

Far January 12, 2013 14:58

All surfaces and edges are recreated in ICEM ;)

https://dl.dropbox.com/u/68746918/project22.tin

Edit : Link will work after about one hour (very very very slow net. With 1 MB connection I am only getting 4 kbs uploaded speed. Shared internet service provider :mad:)

diamondx January 12, 2013 15:01

link is not working or may be upload is not yet done

Far January 13, 2013 11:48

I am working step by step and therefore progress is very slow.

These are my observations :

1. Good geometry is important

2. I need to set the many local parameters along with the curve parameters.

I have one query here:

The common curve between two surfaces should be part of which surface ? Surface 1 or 2? or Both (how is that possible?) ? If I dont build the topology then I have to create the one curve for each surface! In this case how to define the curve parameters?

http://imageshack.us/a/img10/8859/shellmeshing8.png

http://imageshack.us/a/img51/9907/shellmeshing6.png

http://imageshack.us/a/img197/8366/shellmeshing7.png

http://imageshack.us/a/img577/7640/shellmeshing9.png

diamondx January 13, 2013 12:11

Quote:

Good geometry is important
It was easy to generate beautiful surface meshes in your geometry because of the quality of it. your curves and surface are well defined. where did you make them ??
So you are filling up later with volume mesh ?
Tomorrow i will share with you one more complex geometry, not complex but you will see how it's difficult for ICEM to generate a surface mesh while the geometry is a simple cylinder. i guess it because of its quality

Far January 13, 2013 12:17

I had to recreate the each and every point, curve and surface in the ICEM. Before that I had many problems in both methods (octree or surface meshing).

Yes later I will fill the volume with danualy. I am also thinking to take the surface mesh to Tgrid and make the further process there... What do you say?

Far January 13, 2013 12:21

Quote:

Originally Posted by diamondx (Post 401603)
Tomorrow i will share with you one more complex geometry, not complex but you will see how it's difficult for ICEM to generate a surface mesh while the geometry is a simple cylinder. i guess it because of its quality

did you play with the surface mesh parameters? It is very much important to play with global parameters, surface mesh parameters and curve mesh to get the good surface mesh. Things are more complicated in ICEM for some purpose which now I am getting. The purpose to provide more control but it makes things more complicated for novice like me....

diamondx January 13, 2013 12:23

i had never used tgrid !! how about the prism ??? is there a tutorial on how to do that in tgrid ?? is tgrid better than using the tgrid approach when using delaunay in icem ??

diamondx January 13, 2013 12:26

yeah i tried every single parameter from autoblocking to delaunay !!! surface was not successful because of the geometry. you need to define thin cut between the tire and the ground...

Far January 13, 2013 12:29

Quote:

Originally Posted by diamondx (Post 401607)
yeah i tried every single parameter from autoblocking to delaunay !!! surface was not successful because of the geometry. you need to define thin cut between the tire and the ground...

Check out the power of Tgrid here:
https://dl.dropbox.com/u/68746918/AN...st-edition.zip

Yeah I have tried them all. But you need to define many other parameters in global mesh settings, shell meshing setting, surface part mesh setting and ........

No I don't need to define thin cuts for groud and tyre. I need to define thin cuts for the zero thickness plates (which are not shown here)


All times are GMT -4. The time now is 08:03.