Hi

I am new at using Pro-Star and have a problem.

I would like to simulation the flow in two pipes that connect into on pipe formnning a T.

Creating the large pipe is straigforward using the simpel meshing in Prostar, but now I would like to make a small pipe connect onto the middel of the large pipe.

How can I project the small pipe onto the surface of the large pipe.

Any idea ?

Thanks a lot.

Regards

newuser

Short answer: vpro vset ,,,, cset

There a many ways to do this in prostar. Probably this would be a good topic for the adapco-online page.

One example of the correct meshing of a t-junction without singularities -- cells with just an edge or point on the surface -- can be found on

http://www.beilke-cfd.de/t-junction.mdl.gz

but it was not created using prostar

Are you familiar with O-grids? If so, you can use them on your T-pipe model. (Actually, you MUST use O-grids on anything with a circular cross-section or the corner cells of a block will deform badly when they try to conform to the curvature = internal angle -> 180').

I would recommend, before commencing with block meshing with ProSTAR, draw the latticework (er, scaffold for your ProSTAR mesh blocks) in your CAD program first. Good if you have a free-form CAD. Import the latticeworks' splines as an .iges file and you can even overlap the previous surface imported as .stl (STAR-CD sucks at importing .iges). Draw any remaining guide splines across the surface shells and connect them to the splines you previously imported. So, when you finally place the mesh block, their edges will snap nicely onto the splines.

Try visiting:

http://www.icemcfd.com/hexa/ogrid.html

http://www.cham.co.uk/website/new/whatsnew/pic4.gif

(The red lines in the last link would be the latticework I'm talking about).

If you see the picture in the first link you will see that there is no O-Grid within the small pipe. Just a plain h-grid.

The picture at the second link shows the usually used topology for a t-junction. But this is not the "best" topology for this problem (no ogrid around the small pipe).

Turning this topology into the right one just requires one more split ...

Thanks Joern.

The first link is actually an animated .gif picture file which alternates between a H-grid and a O-grid. Between the two, it is easy to guess which is the O-grid by the better cell formation.

There are two ways to block mesh for the geometry in the second link. 1. Hard way: You can either mesh a continuous O-grid (the central 'core' cell block spans from the open end of the large pipe, takes a 90' turn up, and ends at the opening of the small pipe), or, 2. Easy way: you can mesh two O-grids for both pipe segments separately and then later use partial arbitrary couples -the latter is only possible because the curvature of the surface of contact between the large pipe's side and the small pipe's end is not so severe.

I told my browser not to animate any images, thats why i did not see the change to the ogrid .-(

For the ogrid topology: if you look at my example you will see that i used a slightly different topology than that one, which is "normally" used.

To do this you have to split the block of the small pipe near the intersection and move the newly created edges to the intersection of the 2 pipes. The edges, which were originally placed at the intersection are moved a bit outwards on the bigger pipe. So you get also a nice ogrid around the small pipe.

Doing this in prostar is indeed a painful task. With a proper block-mesher it only takes you some minutes.

That's the hard (but best) way.

Even with 'drag-and-drop' block meshers like ICEM-Hexa, pure hex meshing is only for the veterans (you need to spend time to really think about the meshing strategy) and I'd rather just use an automesher like ProAM or ICEM-Tetra for the weird geometries. Those that are familiar with CAD software will find using ICEM-Tetra a breeze and the graphics very much faster. Also, I find that ICEM-Tet has one very useful ability that ProAM lacks: automesh baffles.

It is not hard at all. This example took about 15 minutes to create from scratch (including the geometry).

Most of the meshes I create for my customers have more than a million cells, even with an optimized blockstructured hex mesh. Can you imagine what happens, when the same is done with tets. Thats not funny anymore.

Hmmm, let's say you don't use the same geometry from the ICEM-Hexa tutorials and draw something similar from scratch: set two WCS, draw two cylinders, union both cylinders, draw two cut planes, cut, save as .stl format = 5 mins. Fire-up ICEM-Hexa, open new project, import .stl, extract splines from surfaces, merge splines, cut main block into four sub-blocks, ORFN one sub-block and squeeze the other three into the half-T-pipe, project corners and edges to splines, assign a uniform hexs size value to all surfaces, declare O-grids in all three remaining sub-blocks (terminating surfaces at pipe ends and cut-plane surfaces), adjust O-grid edge cell spacing to taste, and finally calculate projected mesh = 5 to 10 minutes.

ICEM-Tetra will take about the same amount of time for this model because the tet and prism calculations (on a reasonably fast PC) uses the time that Hexa-users take to split, project, adjust blocks. Nice prism layers will help resolve near-wall flows as good as the (naturally) conformal hexa mesh -you can also re-run ICEM-Prism with different cell height values to help adjust for y+ value requirements. Of course, tet meshes take more memory to store and you need more of them to achieve mesh-independant results compared with hexas.

But nowadays I have to rely on ProAM only because my ICEM license has expired some time ago...

...or are you refering to O-grid (or rather, C-grid) block meshing with ProSTAR alone?

I dont use ICEM Hexa simply because it is too bad for my purpose. Gridpro is much more flexible and creates better meshes automatically after creating the topology (you dont have to move the block corners manually as in HEXA).

You only have to logically assign the corners to the surfaces and have to place the near the surface. Then you let gridpro move the corners around to create an optimized mesh.

But Proam is also not too bad And prostar meshing with vgen, cgen ... is also very nice to have a training for your brain (remembering vertex offsets ...).

Yes, I've heard good things about Gridpro too. I wonder if they give out evaluation licenses?

ProAM? I agree with you it's the 'not-too-bad' or 'OK' mesher. But ICEM-Tet still handles better overall, though tetras are not my cells of choice.

ProSTAR? No, it's just too much of an effort to adjust your blocks inside ProSTAR. Better to draw the frame work in a 3D CAD where you can just drag the edges around like ICEM-Hexa. Repeatedly keying in VGEN, CGEN, etc. may also contribute to repetitive stress injury...

Just ask for an evaluation license. Usually they asked for a small fee, but it might have changed.

OK, thanks. I'll try to contact them.

proam hybrid meshes should be able to mesh baffles.

Can you teach me how?

If you use the online user guide in pdf format it tells you how to create hybrid grids. If you then choose the inner mesh (core hexahedral part of the mesh) to exclude a baffle surrounding region, then this region will be meshed with tetrahedral and prismatic cells. And the tetrahedral proam mesher can deal with baffles.

OK, I found a mention of it in the user guide. Thanks I'll try it out. Is this possible for trimmed hex meshes too?

As far as i know one can not use baffles in trimmed meshes. Start with the surface database, then create an inner mesh and classify it. Then one has to choose which part of the inner mesh (hexahedral) should be used as part of the hybrid mesh. Deselect one layer of the inner (hexahedral) mesh and also deselect all cells that are on or touching or are close to the baffle. Then proam will try to bridge the region between the inner (hexahedral) mesh and the surface with pyramids, tetras and a prism layer and it should be able to mesh the baffle. I think you will need to define the lines and points on the baffle and also define the cell type of the baffle to be a baffle. But i remember there are some instructions regarding this in the user manual.