[Daniel C]

Hi,

I am currently investigating an enclosed rotating disc at high Reynolds numbers.

The Fluid volume has a quite simple geometry. It is bordered by two surfaces of revolution, one for the shroud and another for the shaft. There are also three surfaces adjacent to the disc, one for the rim and the others for the disc sides. Finally there are two periodic surfaces which close the volume. You can see the geometrie in the attached pictures below.

I need a node spacing of 0.01 mm on each wall and at this point I got stuck. ICEM found some volume errors and I don't know how to get rid of them. As you can see from the pictures the blocking topologie is also very simple. Just entered 0.01 for spacing 1 and 2 on each edge (except the edges on the periodics) and a ratio of 1.1 . It should be that easy, shouldn't it?

I also checked the associations with the geometrie. I can not find anything. So where am I wrong?

Thank in advance!

[Daniel C]

I saw some skewed elements. I can't fix them.

[Daniel C]

I found a way to resolve this problem. I quit blocking and changed over to a 2D blocking and extrudet it by rotation and loaded the mesh finally with it. So in my opinion it is better to use an unstructured grid for this case.

But I am pretty curious and appriciate every hint, how to mesh this geometry in 3D. Essentially that should be no problem.

[Far]

It is very simple geometry. The possible problems/solutions are:

1. Association

2. Edit edge commands

3. Splitting and snap vertices

[diamondx]

When applying a bunching law, you have to make sure it is applied to the parallel edges as well. but not all the parallel edges has the same direction. may be one of youre edge has the same bunching law but in the reverse direction

[Daniel C]

Thank you Far and diamondx for your hints.

I applied the bunching law to all parallel edges. Yes, sure one has to consider that some edges have reverse spacing direction, but in my case I just apply the same spacing for each direction, so I don't care about the direction.

I loaded a step file from CATIA V5, a very simple geometry. I assigned the outer most vertices to the corner points and split the block twice. Then I deleted the center block which represents the disc material, but not permanently. Finally I assigned all the new vertices to the geometry points and the faces to the respective surfaces. I controlled everything by EDGE\Show Association. All vertices are aligned and associated with the geometry.

I did it over and over again, but could not get any utilizable result. But if I use an unstructured 2D mesh, and let ICEM extrude it by rotation, everything is fine.

If it is not caused by my step file, I would say ICEM is very buggy.

[Daniel C]

Here is my tin file, just if somebody is interested in resolving my problem.

[Far]

I didn't find any problem.

Can you attach the problematic blocking?

[Daniel C]

Here is my blocking.

Thank you in advance!

[Far]

Is it necessary to use .01? Why not .03?

[Daniel C]

Far, I used your blocking and got the same problems again. There are very few cells, what are erroneous. They should be identical to the other good cells, because of the constant curvature.

I am obviously doing something wrong.

[Far]

Those cells are not perfect rectangles as other cells are. Dont know why !!!

[Daniel C]

It is not necessary to use exactly 0,01 but I need to refine the wall spacing to get a certain y+ for the wall function. In the worst case I have to go below 0.01, namely 0,001.

It should be possible with 0,01, shouldn't it? Or is ICEM not capable to build such a grid with the blocking approach?

As I said, my problem is resolved. I use a 2D mesh and convert it to an unstructured mesh. The circumferential elongation is achieved by extruding the 2D mesh by rotation. I am usung -1° angle per layer and using 1 layer to get a 15° segment.

[Far]

Hey I found the solution. It was a geometry problem. Deleted all surfaces and recreated all from curves.

I even used 0.001 without any problem.

Although I made the blocking for one portion of the geometry this time but it is enough to prove my proposition.

As I said, my problem is resolved. I use a 2D mesh and convert it to an unstructured mesh. The circumferential elongation is achieved by extruding the 2D mesh by rotation. I am using -1° angle per layer and using 1 layer to get a 15° segment.

Because here your mesh is not being influenced by the geometry inaccuracies.

[Daniel C]

Far you are great!

That is a very subtle geometry failure. I will carefully reviev any geomery that is derived from a step file in the future.

I always appreciate your efforts!

[Far]

Check this and give some comments that what is happening in ICEM

[Daniel C]

Haha, you have an accurate eye

After you discovered the issue with the erroneous surfaces, it was clear to me that this must be the source of error. So I tried to resolve it according to your advice and got rid of the problem. But I checked your solution and it is fine. So thank you very much Far, and once again I appreciate your efforts.

[Far]

Actually it is not only the problem due to the erroneous geometry only. Say it is 30%.

You can observe I have made the intermediate surfaces between the thin and large blocks on both sides. First to get the overall idea I only made that surface in one portion. And to my surprise I got problem in the other side (where surface is not present) although with lower intensity.

So what I conclude that closed volume is necessary to get the prefect rectangles.

Also to avoid the walls due to those internal surfaces I placed the materials points with same names in all those closed surfaces.

Got it!!! Life is full of trial and error