[RodriguezFatz]

Dear all,

I am meshing a pipe of diameter 25e-3m. For high velocities I need a y spacing at the wall of 1e-6m. Now in ICEM I have a problem that I am not able to solve. I don't get "negative volume" cells at the wall (like in other threads), but just cells with extremely bad orthogonal quality (0-0.05).
ICEM_ortho.png
I already played with the "projection limit" and all that stuff, but that doesn't help. It seems like my geometry is somehow crappy!? I did it with DesignModeler.
Does anyone know how to fix this?

[kad]

Somehow I can not reproduce this is error in ICEM 14.5 . Geometry and blocking are attached. Geometry was made in ICEM and blocking was done with default settings. Did you try to lower the tolerances under settings -> model, too?

[Far]

with 0.1 tolerance, mesh is not following circular shape of cylinder, but still orthogonal quality is at least 0.5.

With standard setting, I was able to use 1e-7 spacing and still getting quality of 0.5 angle > 45.

Geometry was made in spaceclaim.

[RodriguezFatz]

Hey guys, sorry for the delay:
Thanks for your answers. I didn't mention, because I didn't think it's important:
I have a bluff body (for vortex street) inside the pipe (see below). I attached my blocking.
Now the important part: I only see the distrortions in "orthogonal quality" after converting the grid to unstructured. The other qualities (determinant, ...) are quite good as well.

vortex.zip

[kad]

I still get orthogonal quality >0.2 which isn't actually that good. But the bad quality elements are due to improper blocking and not to an error or something.

[RodriguezFatz]

Hey kad and far, I really appreciate your help, thanks! So it looks like that the same blocking leads to strange boundary quality in my mesh in opposition to kad's. This is weird. So basically any of my "general" ICEM settings is messed up? Does anyone have an idea which one? Tolerance, Topo tolerance and all that stuff?
Probably the easiest way is just to install a clean copy of ICEM...

[Far]

Before reinstalling ICEM CFD Try "setting > Restore & Reset"

[RodriguezFatz]

Yes, I just did that. This is how it looks like.

I'm doing: Pre-mesh-Recompute, convert to unstructured mesh, Edit-Mesh->Display Quality-> Orthogonal Quality.


ortho.jpg
model.png
hexa.jpg

[Far]

Min ortho quality > = 0.6

Min Angle > = 27

Am I missing something

[RodriguezFatz]

Far, the "Determinant 2x2x2" is fine in my case, too. It is the "orthogonal quality" of the unstructured mesh that is bad.

[RodriguezFatz]

Ok guys, mea culpa.
I just opened the same case in ICEM 14.5 which I kept installed based on laziness, and tadaaa: It works!
I try to figure out why this happens in 15.0 and will post it.
Thanks again.

[Far]

so you have just found a bug in ICEM CFD 15.0.

I am still not able to understand if every thing is ok (angle , determinant) why ortho quality is going down?

[RodriguezFatz]

Something is really wrong with the computation... I create the unstructured mesh in ICEM 14.5. Everything is fine, worst cells are orthognoal quality >0.2 and due to the crap I did with the O-Grid of the bluff-body.
Now I just load the .uns file in ICEM 15.0 and without changig anything orthogonal quality goes to 0.1 and worst cells are those boundary cells.
I'm going to write to ANSYS and see if they are aware of this, of if I do something wrong.


Edit: I just wrote a service request. I will keep you informed.

[Far]

Originally Posted by RodriguezFatz

Something is really wrong with the computation... I create the unstructured mesh in ICEM 14.5. Everything is fine, worst cells are orthognoal quality >0.2 and due to the crap I did with the O-Grid of the bluff-body.
Now I just load the .uns file in ICEM 15.0 and without changig anything orthogonal quality goes to 0.1 and worst cells are those boundary cells.
I'm going to write to ANSYS and see if they are aware of this, of if I do something wrong.


Edit: I just wrote a service request. I will keep you informed.

it may not be the only computation problem. Check both meshes in Fluent and compare orthogonal quality there as well.

[RodriguezFatz]

Nice idea. I checked the meshes and it looks like that ICEM 15.0, Fluent 15.0, 14.5 and 14.0 all say the mesh is bad (min. orth. qu. = 0.124512), only ICEM 14.5 says it is much better.

[kad]

That's interesting. I also have V14.5 and 15.0 installed and I will try if I get the same results.

Okay, I can confirm the error in V15.0. That is really a bit strange. You can dimish this effect by adding more ogrid nodes.

[RodriguezFatz]

I thought about the reason for this. If these cells are actually that bad, why would all other quality criteria say, they are good? Can you artificially create cells with good skewness, determinant qu., ... and bad orthogonal quality?
Also, the edges in "y-direction" close to the wall are more or less straight lines. So if you cut one of the boundary cells into more cells to refine the grid, the cells will keep their shape (except the width in y-direction). Why does the orthogonal quality decrease?
Maybe it is just some artifact in the calculation for very thin cells and it's meaningless?

[kad]

Originally Posted by RodriguezFatz

So if you cut one of the boundary cells into more cells to refine the grid, the cells will keep their shape (except the width in y-direction) Why does the orthogonal quality decrease?
Maybe it is just some artifact in the calculation for very thin cells and it's meaningless?

I made a little drawing for this. I hope I got it right. Look at the definition of orthogonal quality in ICEM. It calculates the cosine between Af/C and Af/F. The minimum value is used. When decreasing the number of nodes the cell centers move down. This causes the angle between Af/F to become larger, so it's cosine becomes smaller.

The manual for V14.5 states that the calculation procedure in ICEM, Fluent and CFX are in general the same, but there can be differences due to different tolerances used by the programs.

[RodriguezFatz]

Just to get the sketch right: The wall is at the bottom, only the wall cell is shown?

The cell centers move down... but also the face centers. Doesn't the angle keep constant?
Even if not: At the pipe wall there is no bend like in the picture. These cells should be nearly 100% orthogonal.

[kad]

The two cells are an example for the very first cells away from the wall. So the pipe wall is at the top. Let's assume the verts of the connecting face are on a radial ogrid edge so they definitely stay in place. Then of course the face does not move too while the adjacent cell centers are going down with decreasing node number. Btw. this is exactly the region where the low quality cells appear in yor mesh.

Of course this is just an attemp to explain the behaviour. But it could be possible in addition with some change in tolerances from 14.5 to 15.0 when evaluating the metric.

The near wall cells are orthogonal to the pipe wall so their angles differ from 90 degree. Okay, the edges in the red circle are not really orthogonal to the pipe wall, but they are supposed to be as it was areal quick drawing.