[sokal_]

Hello,
Can anyone help me with my problem.
I struggle to get a good quality hex mesh on a cyclone. (fig.) I've got nearly flat elements on a tangential connection (fig.).
How to avoid that?

[sokal_]

Please guys,
Every second deals with cyclones. Share your knowledge.

[bharath]

hi sokal,
i have a suggestion on this.you split the inlet portion from cylindrical portion with little protrusion of inlet from other one and vice versa for the other.

[sokal_]

thanks for your reply bharath,
but I don't really understand what you mean. Can you draw it?
I've heard that people do something with blocks vertices, but what?

[bharath]

hi sokal,
try to split your domain into three parts like in this attachment and mesh them separately then make interfaces. http://img13.imageshack.us/img13/4475/splitgeometry.jpg

[bharath]

but take care on mesh

[PSYMN]

I would use a quarter ogrid (AKA Y-Block) topology on that "triangle" section. Mesh the rest of the cylinder alone and the botom up create three blocks in this configuration... Actually, I took a screen shot, but don't have time to figure out how to post the image on this site (turn it into a URL...)

The triangle is made up of three blocks each with a corner on the three corners of your big triangle and the oposite block corners are in the middle of the wedge. The remaining corners are all about half way down the sides of the wedge...

This will leave you with some farily accute angles in the tangential corners, you you can run a final ogrid thru the model to solve that problem...

As I said, I have the images, if you can tell me how to post them, I will.

[sokal_]

well, split triangle from the rest of cylindre is one of the ways but I'm trying to avoid that. The model I've attached a picture of is a simplificaton of a real one. The real one contains 4 triangles such that on different angles along a cylinder. But anyway I'll have to do that if I won't be able to find another way within next several days
Thanks for suggestion bharath.
2 PSYMN
If you mean what I think about then I have probaly tried that
The problem with Y-Block was in connecting a part with cyclinder+triangle to pure cylinders on left and right sides.
About a picture. Instead of using quck reply click "post reply" scroll a page down and you will see "Manage Attachements" in "additional options". You can upload pictures there.
I'll be waiting for your pictures.

[PSYMN]

Here are the pics...

Another option, if your solver doesn't insist on multi-block, is to use the new swept blocks in ICEM CFD.

[jsm]

Hi,

Cyclone geometry can be modifed slightly. The inlet duct can be offset with 0.1 or 0.2 mm instead of tangential connection to cyclone body where highly skewed elements are coming. It will not change much difference in the simulation results. This is just one suggestion can be used.

with regards,
JSM

[sokal_]

Thanks jsm
I've a thought about that but saw that people bypass that. But it's still a method and it's pretty quick method.
2 PSYMN
Your pictures look great!
Looks like it's what I'm looking for.
How did you manage to split blocks 24 and 35 to have 63 and 64? By using O-grid? If yes then which sides of blocks did you choose for O-grid? because I can't get the same result! and if you have time could you please write some major steps you take to get that nice looking grid.

[PSYMN]

That is just the automatic OGRID function...

Block it out as in the first picture... You may need to use bottom up YBlock to select the six corners... Or you could do it top down if you know how.

Then Blocking => Split Blocks => OGrid.

Select all the blocks.

Select the flat faces at the ends of the big cylinders (Optional)

Apply.

[bharath]

Hi sokal,
have u tried simon's suggestion in your mesh?is it working well?

[sokal_]

well, if to be honest I gave up after spending several days trying to get what Simon has suggested. Everything is fine when you consider only a part with a tangential connection, but what about two cylinders on both sides of that?!? Have a look on Simon's mesh and move it forwad in your mind to the point where tangential connection finishes and a pure cylinder begins (my first pic. might help), what happens to mesh over there?
Nevertheless splitting geometry to pure parts and parts with tangential connections and using interfaces between them or make an offset of 1-2 mm as jsm has offerd work fine.
If you have another solution I'll be pleased to know that.

[PSYMN]

Hello,

I took a few mintues to block this out.

I started with a cylinder OGrid, then extruded the two faces and then built the last corner block from 8 verts (6 blocking and the final 2 on the geometry). Then I ran Ogrid with faces on the ends and central pipe. I had to clean up the boundary layer a bit because I had set it to larger than the corner block size. I just used Set location to clean it up. When i meshed it, I noticed that the internal quality was lower because the internal edges were bending to match the surface, so I straightend those out with linear edge splits... Total blocking time ~ 5 minutes, it took longer to upload and write this email ;^).

If I had time, I could go further to improve quality, maybe adjust the size of that corner block. Violating some of the geometry at the tangent would make the biggest difference.

Currently, Min Determinant is 0.31. But angles are probably worse, I didn't take the time for an angle check but I checked the ICEM CFD catch all quality and am including that image also...

The ICEM CFD files are in a zip folder attached.

[arjun3020]

Hi SPYMN,

I want to use the quarter o-Grid, i have done few steps. but i cant proceed. i have attached picture for your reference.
now while creating y block (quarter o- grid) it ask to select 6 vertices. which vertices should i select. also tell me about association i have to do after creating Y block. please help me.