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] Creating prism layer for a badminton shuttlecock model decreases quality (http://www.cfd-online.com/Forums/ansys-meshing/95707-creating-prism-layer-badminton-shuttlecock-model-decreases-quality.html)

li0012in December 29, 2011 10:00

Creating prism layer for a badminton shuttlecock model decreases quality
 
2 Attachment(s)
I am trying to mesh a badminton shuttlecock within a fluid. That means, I have it like how the "blunt body tutorial" is teaching- a shuttlecock "boxed" by a liquid body.

My problem is that with tetra only, the mesh quality is relatively acceptable. When prism layer is added, the quality goes all the way down. This is despite a reduced maximum size of the prism layer.

I have attached photos.

The problem I have is that the feather of the shuttlecock overlaps each other. In between the overlap of two feathers is a small tight sharp corner. At that area, the mesh elements are of very poor quality, even at decreased size.

I can't quite decrease the size too much because too many elements will cause my computer to run out of memory when loading the prism.uns thingy. Yet, my computer is a 2600k with 16gb of ram...so I really think the problem is ME, not hardware limitation. I mean this is just a shuttlecock, can't be that complicated right?

1. What should I do to improve the mesh quality?

2. Since shuttle is within a fluid, I should set the shuttlecock surface as split wall?

li0012in December 29, 2011 10:52

Also, how can I set the prism density. Setting a density through block or by creating density box around entity does not seem to have effect on prism element?

PSYMN December 30, 2011 15:42

Yea, that tight area is going to give prisms a problem...

I would close it off so that the prism went around it, but put another material point inside each of those areas so you still get fluid in there.

After meshing and adding your prism layer, delete the shells from the construction surface part and put all the volume elements in the same part.

Make sure that when you setup your prism, you set the prism to grow in the outer volume part only and not in the slivers.

As for setting prism sizes, you can do that thru global settings or on individual entities. If you want it to be controlled by your density box, be sure not to set the initial height or total height. Then prism will fluctuate the initial height such that the last height matches the volume of the adjacent tetrahedral (extrapolated from the size of the surface triangle)

PSYMN December 30, 2011 15:43

One other thing... Even though you won't have symmetrical boundary conditions, you could mesh just one segment of this model (with periodicity) and then copy/rotate it around.

li0012in December 31, 2011 05:36

Thank you for the reply.

I don't quite understand

Quote:

I would close it off so that the prism went around it, but put another material point inside each of those areas so you still get fluid in there.

After meshing and adding your prism layer, delete the shells from the construction surface part and put all the volume elements in the same part.

Make sure that when you setup your prism, you set the prism to grow in the outer volume part only and not in the slivers.
Does it mean that I...

1. Create the inlet, outlet and wall boundary parts as usual, but not the shuttlecock model wall.

2. Create surfaces to close up the sharp corners in the feather-feather interaction areas. That means if I have 10 sharp corners, I will have 10 newly created surfaces to cover up the corners for meshing of prism?

3. Add new part of shuttlecock model wall by selecting the surfaces of the original shuttlecock model with the newly created surface for closing the gap, less the sharp corner feather-feather overlap area.

4. Create a shell around that new part.

5. Create another body of fluid for every volume that is covered by the sharp corner and the newly created surface from step 2. That means I create 1x body for each region. So if I have 10 newly created surfaces, then I will have 10 new body of fluid+1 original body of fluid in the large box?

In this case, do I need to specify anything to tell the software that the bodies are "linked"?

I don't quite understand the part about deleting the shell. With the newly created surfaces to block off the sharp corners, my shell will be created around the new surfaces. If I delete the shell at some places, wouldn't the shell be incomplete?

li0012in December 31, 2011 18:31

Actually...I was thinking about it the whole night of new year eve(I am at gmt+8)...what a way to start a new year;)

Do you think it will work if I mesh it without the prism layer using the default shuttlecock model with all the sharp edges.

Then cover up the sharp corners of feather-feather interaction using "extending of curve to a surface function", to create a surface from the trailing edge of a feather direct to the surface of the feather that is overlaps.

Then use that model to mesh the prism layer only.

In that case all the "fluid" will still be connected.

Do you think this method is sound?

PSYMN January 1, 2012 22:40

More Details...
 
Nope, the simplest plan is just to create the construction geometry, mesh it, then delete the construction triangles.

Here is a bit more detail...

The problem areas occur where the feathers overlap... These feathers overlap to form an "eye" shaped region. What we need to do is close that off temporarily with a construction surface. Put material points in the void. Then tetra mesh the whole model (all the voids and the major flow volume). This will ensure connectivity between all the regions. If you use the octree method, it will generate the volume and surface mesh all at once. You may need to use finer mesh and/or apply "thin cuts" in the tight regions, but I guess you have already figured out how to get a mesh. Then when you prism mesh, grow prism on the surfaces (including the construction surfaces), but into the main volume only (not any side volumes). At the end of the prism step, you will have a main volume (FLUID) containing prisms and tetras. You will also have all the smaller volumes, each containing only tetras in some other volume part (NAME_IT_WHATEVER). Between the fluids will be the shells of the construction part. So you will need to right click on FLUID in the tree and select add to part. Then select all the volume elements in the model (there is a one click button in the selection tool bar for this). Then you will need to delete the construction elements (select all elements in a part is also an option in the selection tool bar).

But lets back up a step.

Your model is periodic (15 feathers)... I know it will eventually be in a larger space at some strange angle, but for meshing purposes we can work with submodels. I would suggest to cut out the smallest portion that is repeatable (1/15th). Put it in a wedge of a cylinder designed with a nice nominal size. Setup the periodic angle under Mesh => Global Settings. Later, after meshing that segment, you can copy it around and then put the full cylinder sub model into a larger model... You can then rotate it or what ever you need. I really recommend meshing just a small periodic section because you need a fine mesh in those gaps. If you don't have more than one element across a gap, the solver can't do much, and it is easier to handle a periodic portion of a mesh than handling the fine mesh all at once.

To create the construction surfaces, I would take the curves on the edge of one feather and project them to the surface of the nearest feather, then create a surface from 2 curves stretched between each source and target pair until you close the gap...

Any other specific questions?

Happy New Year!

Far January 2, 2012 07:40

Quote:

I would close it off so that the prism went around it, but put another material point inside each of those areas so you still get fluid in there.
Why it is necessary to put another material point? material point is equivalent to default interior in fluent right?

PSYMN January 2, 2012 09:17

There are a number of reasons...

First, if you didn't put separate material points, the construction surfaces would block the mesh and you would get nothing in those overlapping feather regions.

Second, if you put the same material point in those regions you wouldn't get any construction triangles to grow the prisms off... ICEM CFD octree assumes that if you have the same region on either side, you don't want a wall between (unless you specify it as an internal wall).

Third, even if you did specify the construction family as an internal wall so you could skip the step of putting all the material points in the same part from the start, it is only with a separate part name for the main volume that I can control prism to grow in only one direction from the construction surface...

Far January 2, 2012 10:12

Simon you are a great help and inspiration to this forum.

li0012in January 5, 2012 01:36

Thank you for the great advice.

Will need to try it out now...


All times are GMT -4. The time now is 22:55.