unsatisfying meshing behavior of a spherical shell
unsatisfying meshing behavior of a spherical shellPosted by pierre_j at August 22. 2012
I created a spherical shell mesh (1/8 of a sphere actually) that I took great care to partition correctly in order to ease meshing.
(I am far from being skilled: there might be far easier approach to get the partition, but at least, it has been done)
I then meshed this 8th of a sphere.
You will find enclosed the python script that I cleaned.
What I find unsatisfying in the result mesh is the top third of the mesh, and most particularly the top element that is larger than the other elements when one might expect equivalent size with the 2 other extreme elements (with respect to X and Z axis).
This expectation might seem somewhat strict, but then, let's turn the mesh to quadratic with mid nodes on geometry: this element simply turns ugly.
I don't think that this behaviour is acceptable, and wanted to report it. However, I don't have any idea where might be the mistake.
I mean, the way it is partitioned, SMESH shouldn't behave like this, should it?
I thank you for your comment to improve this simple mesh, because I actually would like to use it .
Have a good evening.
PS: using Salome 6.5.0
Re: unsatisfying meshing behavior of a spherical shellPosted by pierre_j at August 22. 2012
Here is a picture highlighting the element that I find ugly after having turned it to quadratic.
Re: unsatisfying meshing behavior of a spherical shellPosted by SMESH expert at August 22. 2012
This case is very specific because actually the face, mesh on which you complain about, has 5 edges and not 4. The 5-th edge is a degenerated edge at sphere's apex. In parametric 2D space this face looks like a pentagon. It required many tricks (including smoothing) to make such faces meshed with quadrangles of this quality if ever.
To avoid this I advise you to cut off a part of the sphere that does not include the apex (see your modified script).
Re: unsatisfying meshing behavior of a spherical shellPosted by pierre_j at August 23. 2012
Thank you very much for your help.
I understand your explanation.
I understand that such a degenerated edge might be needed to represent the shape in GEOM and therefore understand that it cannot be solved with the healing procedure that should return a geometrical shape.
On the other hand, I am surprised that this could not be solved by a pre-treatment procedure in SMESH checking for degenerated edge, and merging of nearly coincident nodes just before meshing. Such a procedure could be turned on by a boolean just before meshing.
Is it not a viable solution, far less complex than post treatment of mesh with smoothing procedure?
Re: unsatisfying meshing behavior of a spherical shellPosted by SMESH expert at August 23. 2012
The mapping algorithm used by "Quadrangle (mapping)" mesher works in 2D space and 2D result is then mapped into 3D. Thus the issue is that the face is 5-sided in 2D. The attached picture shows it's 2D representation. The upper horizontal edge there is one at sphere apex.
So a current solution includes not only smoothing.
Re: unsatisfying meshing behavior of a spherical shellPosted by SMESH expert at August 31. 2015
I finally found a way to fix your initial case. Here is a quadratic and a liner mesh after the fix:
Woah, that is support!
Edward, please, how should this be handled? (what is the trick?)
I thank you in advance for your answer.
Re: unsatisfying meshing behavior of a spherical shellPosted by SMESH expert at September 01. 2015
I modified the algorithm of Quadrangle (Mapping) to use TransFinite Interpolation in 3D to find positions of new nodes in such a case, instead of ( pretreatment -> TFI in 2D -> mapping into 3D -> smoothing ).
The fix will be available in v7.7.0.
Great news! Thanks for sharing it!