CFD Online Discussion Forums

CFD Online Discussion Forums (
-   ANSYS Meshing & Geometry (
-   -   [ICEM] Blocking strategy around a sphere (

Mudblood May 17, 2010 15:58

Blocking strategy around a sphere
1 Attachment(s)
I learned how to block around a circle in 2D, and now I'm trying to learn how to do blocking in 3D. I followed the Sphere-cube tutorial in the ANSYS customer portal. I tried to do similar around a sphere but did not have a good result.

I'm trying to do flow past a sphere, so i did an OGRID and then tried a c-block strategy similar to PSYMN's airfoil tutorial on youtube.


PSYMN May 18, 2010 08:59

The first of the three parts in that youtube video (the ppt section) showed some far fields... We used a CGrid for the demo because the far field was shaped that way. In this case, your far field is a box, so an HGrid out there would be fine. Putting in the CGrid when your model has corners like that actually reduces quality in the corners. On the other hand, perhaps you were looking for the refinement like that and should change your far field shape instead...

Assuming you decide to keep the box far field... Here are the suggested steps.

Start by initializing blocking and then use the auto associate (last icon under associate) to automatically associate to the box. Then create 2 splits, just a 1/3 diameter on either side of the sphere in each direction (6 splits total). This should form a box around your sphere. Use the index control (right click on the blocking branch of the tree or use the "i" hotkey while in the blocking tab) and adjust it until you only see that inner box. Use Split Block => OGrid and select just that center block. No need for faces since we want the ogrid to go all the way around. Delete the inside of the created ogrid (assuming you want a hollow sphere) so that the inside of the Ogrid default associates with the nearest surface. Use the Snap Project Verts button (middle of the second row of associate DEZ). This completes the blocking and association... Now you can setup edge params, etc...

As a side note, OGrids are not really needed to improve mesh quality for the outside of curves. An HGrid will fit to the outside of a sphere without giving bad quality elements. However, it is still good to have in this situation because we can use it to generate a nice boundary layer grid (assuming you are looking at viscous flow around this sphere).

Best regards,


Mudblood May 18, 2010 13:36

Thanks for your help! got it to work =)

lost.identity April 7, 2011 07:34

Hi, I'm also trying to do a sphere inside a rectangular box. However, I'm only considering 1/4 domain. I will be doing a combustion simulation with the sphere denoting the initial spherical flame which would be convected downstream. I'm not sure how I would go about doing this using an O-grid. The sphere is very small and coloured in green in the image (on the left hand edge).

PSYMN April 8, 2011 09:08

Faces on Symmetry Planes...
Your image didn't make it, but I guess you are talking about a quarter symmetry model...

The steps are all the same, except that you should apply "faces" to the symmetry planes when creating oGrids...

lost.identity April 8, 2011 13:06

1 Attachment(s)
Thanks for the reply Simon. I followed the instructions you gave earlier, my problem is the edges doesn't associate to the curve of the sphere when I click the 'Snap Project Vertices' button. I've attached a picture of what it looks like when I mesh. What am I doing wrong?

PSYMN April 8, 2011 21:41

Edge to Curve
They are projecting to the nearest surface, which must be the side walls...

Presumably you have a curve at the intersection of the sphere and the symmetry planes... Associate the appropriate edges to those...

In fact, where ever you have a curve, you should always associate the edges, even if it doesn't look bad right now (such as the straight sides of the box)...

Best regards,


lost.identity April 9, 2011 05:54

1 Attachment(s)
Thanks for the reply. I observe the same thing when I try to do the full sphere, here I'm trying to follow what you described exactly. The snappy project vertices doesn't seem to do anything.

PSYMN April 9, 2011 10:30

??? Snap project verts simply projects the verts to the nearest location on what they are projected to. In your case, they are snapping to the nearest surface.

It worked in your quarter symmetry model, its just that the nearest surface was the side walls.

Instead, associate that edge to curve and you will be all set.

All times are GMT -4. The time now is 08:35.