How to mesh above an existing manually generated mesh ?...

Hi all

For a very specific situation, I need to create a part of a 3D mesh manually (with the Revolution tool) and to mesh the rest of the mesh with algorithms (Netgen, Hexahedron (i,j,k), 3D Extrusion, ...).

But I have not been able to do this...

I have tried on a very simple 'test' case. I want to split a mesh in two parts: one meshed manually and the other with algorithms.

I tried the following strategy but it fails:

• - In the geometry, create two volume groups having the shape of the different areas (to mesh manually and with algorithms)
• - Create a void mesh
• - Mesh a face
• - Make a revolution of this face to fill the first volume geometrical group
• - Create a mesh group including the resulting 2D elements of this mesh on the rest of this revolution
• - Create a submesh with the 2D algorithm Use Existing 2D Elements (What is the difference with Use Existing Faces?) to 'copy' the manually generated 2D element on the submesh.
• - Here, I would like to mesh the other geometrical volume group with algorithm, but I don't know why, some elements (3D and 2D) of my manually generated mesh are deleted...

Difficult to be clear just by words...

Enclosed is a dump study of this case... I made the revolution... Do anybody knows how to mesh the rest of the mesh with, let's say, the Netgen algorithm? (pyramids and tetrahedrons)

Thank you very much

Best regards,

William

Actually, my very first problem is that I have to fill the 2D mesh enclosed with prisms, and the 3D extrusion algorithm doens't know which face to extrude...

Error message:

Invalid input mesh. More than 2 faces with not quadrangle elements: 4

So I concluded that I had to fill it by myself with a revolution. But how to include this mesh in the rest of the mesh?

If anyone has an idea how to do it... it would be very greate...

William

HI William

I have not yet read your second post in this thread, I've just won cylinder.py from your first post (see attachment with comments enclosed).

It seems that cylinder_FIXED.py answers to your second post as well.

St.Michael

Hi Saint Michael

Thank you very much, it is exactly what I need

I will study your script with greate pleasure

Unfortunatelly, I have another problem after exporting this mesh in Edge format (it finds a deviation of accumulated surface vector extremly high at a point laying on the contact between the 'manual mesh' and the 'algorithm mesh'...), but I think nobody here can help me with this...

Thank you very much !

William

It works !

I simply made a mistake during the boundary group definition

So I will be able to use your solution, thank you very much !

William

Hello, Saint Michael

There is a little problem with your solution... Because it requires to mesh the 2D "interface" between the 'manual mesh' and the 'algorithm mesh' in advance. But the problem is that in the case of the sphere quarter I want to mesh in fine (cf. the last picture I enclosed), this is not possible...

Indeed, if I try to mesh the interface with the Quadrangles algorithm, and then if I create the revolution, the nodes doesn't match at all (see a screenshot here : http://nsm08.casimages.com/img/2012/12/20/12122009232114358110684049.png).

That's why I wanted to begin with the revolution, and then with the algorithms...

Don't you know a way to do it like this?...

I enclose a dump of the 'test' study used to create the screenshot.

William

Hello William

Maybe it's a solution to use a larger tolerance at merging nodes of rotation mesh with the rest mesh? Please consider my changes in your script.

St.Michael

Hello Saint Micheal

This solution fits only for very coarse meshes...

On the enclosed picture your can see pairs of nodes which have to be fused on a fine mesh and which are obviously too much far each from the other to be fused correctly...

I think it should be a way to do it more... properly ! Some kind of generic method which can be used even for fine meshes.

Maybe by using two meshes instead of one, and to use Existing 2D Elements on the interface surface from one mesh to another...

I will make some tests and report the results.

William

OK

I did it successfully by this way:

• 1- Create a void mesh
• 2- Create in this mesh only the interface between the 'manual mesh' and the 'algorithm mesh' thanks to a 2D revolution
• 3- Create another void mesh (which will be the 'final' mesh)
• 4- 'Import' in this mesh the interface thanks to an Use Existing 2D Elements algorithm from the first mesh
• 5- Mesh the final mesh with algorithms
• 6- Finally, make the 3D revolution in the final mesh
• (7- Of course, merge the nodes and the 2D elements)

This is not very elegant because it requires to make two revolutions instead of one, but it seems to work fine !

Thank you Saint Michael for your help. You learned me about the multi-dimensional nature of Use Existing Elements (this explained some troubles I had before), and about the element fusion

Thank for all your time and Happy Christmas !

Best regards,

William