[ICEM] meshing a sphere - large deviation from perfect sphere

mjgraf · August 14, 2012, 10:52

wanted to mention, when running CFX (most likely other solvers too), best practice is to run double precision with y+ <= 1
try this with CFX and see if your problem remains. If you are already running double, hmmm.

flotus1 · August 14, 2012, 11:14

This is correct for turbulent flows, where Yplus<1 can (usually) only be achieved with very thin cells (high aspect ratio) in the boundary layer.
In this case, I couldn't find any difference between single and double precision.

mjgraf · August 14, 2012, 11:46

reread, this is laminar flow.
interesting.

murx · August 15, 2012, 07:59

Yes it is laminar. The y+ value is very well below 1.
Double precision also did not make any difference in my simulations.

flotus1 sent me his fluent file. Unfortunately, I was able to discover a similar mesh-imprint in the pressure profile as in my CFX results. The attached picture shows the pressure profile obtained from the fluent run.

I know that this effect is not very strong. But since I my goal is to investigate the forces acting on the sphere, this is something I cannot ignore.

flotus1 · August 15, 2012, 12:53

As I said already, the mesh should be created with a lot more patience than I had. In my file, the "imprint" doesn't come from the irregular radius of the sphere, but from the size and angle change between two blocks. This effect can be minimized with a proper mesh, but will never vanish completely.

The only other choice you have are tet or poly meshes. Here you won't see any imprint from the mesh. Nevertheless, the results will be less accurate than with a hex mesh.

murx · August 15, 2012, 13:37

Yes, I guess you are right.

Although the irregularities only occour directly on the sphere surface and not along the whole block-edge and only when the sphere is rotating. But still... I also do believe it is the mesh. With the new blocking approach, mesh optimization reduced the irregularities a little.

So... thanks for the help. Case closed

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Meshing a Sphere	Ajay	FLUENT	10	September 3, 2016 15:18
[snappyHexMesh] Meshing a sphere with snappyHexMesh	Cyberholmes	OpenFOAM Meshing & Mesh Conversion	2	July 19, 2011 17:46
Meshing a sphere using ICEM	lost.identity	ANSYS Meshing & Geometry	1	March 13, 2011 21:19
Large U-tube modeing and meshing.	Pradeep	FLUENT	1	April 17, 2007 04:08
meshing F1 front wing	Steve	FLUENT	0	April 17, 2003 13:37

August 14, 2012, 10:52		#21
mjgraf Senior Member Join Date: Dec 2009 Posts: 131 Rep Power: 19	wanted to mention, when running CFX (most likely other solvers too), best practice is to run double precision with y+ <= 1 try this with CFX and see if your problem remains. If you are already running double, hmmm.

August 14, 2012, 11:14		#22
flotus1 Super Moderator Alex Join Date: Jun 2012 Location: Germany Posts: 3,427 Rep Power: 49	This is correct for turbulent flows, where Yplus<1 can (usually) only be achieved with very thin cells (high aspect ratio) in the boundary layer. In this case, I couldn't find any difference between single and double precision.

August 14, 2012, 11:46		#23
mjgraf Senior Member Join Date: Dec 2009 Posts: 131 Rep Power: 19	reread, this is laminar flow. interesting.

August 15, 2012, 12:53		#25
flotus1 Super Moderator Alex Join Date: Jun 2012 Location: Germany Posts: 3,427 Rep Power: 49	As I said already, the mesh should be created with a lot more patience than I had. In my file, the "imprint" doesn't come from the irregular radius of the sphere, but from the size and angle change between two blocks. This effect can be minimized with a proper mesh, but will never vanish completely. The only other choice you have are tet or poly meshes. Here you won't see any imprint from the mesh. Nevertheless, the results will be less accurate than with a hex mesh.

August 15, 2012, 13:37		#26
murx Member Max Join Date: May 2011 Location: old europe Posts: 88 Rep Power: 15	Yes, I guess you are right. Although the irregularities only occour directly on the sphere surface and not along the whole block-edge and only when the sphere is rotating. But still... I also do believe it is the mesh. With the new blocking approach, mesh optimization reduced the irregularities a little. So... thanks for the help. Case closed