Page 1 of 2

prism+tets

Posted: Mon Nov 02, 2015 5:18 pm
by ewpostek
Dear Colleagues,

I do an attempt to generate a single layer mesh of 6-nodes prisms
on an unstructured tet-mesh. This is to have single layer
of shb-type shell elements covering the tetrahedral mesh.
Therefore, I generated a quarter of a sphere (volume to be
meshed with the first order tets) an a "skin" of a 3d-layer
(volume to be meshed with the 6-nodes prisms).

However, I am not successful at all, but it looks
it should be possible.
Could give any advice ?

Best regards,
Elek Postek

Re: prism+tets

Posted: Mon Nov 02, 2015 6:19 pm
by abelcs
Hello,

the prism elements can only be obtained from a topologically prismatic volume. Can you send me the model in order to allow me to check whether the geometry fits with this requirements or not?

Re: prism+tets

Posted: Mon Nov 02, 2015 6:22 pm
by escolano
If do you already have created the 'ring volume' to be meshed with prisms, and this volume is topologically two similar tops, connected by lateral four-sided surfaces,
(you can create this volume for example form the copy window, with a offset of surfaces extruding a volume, or in the particular case of a sphere instead an offset operation a scale could be used. Offset is an approximated operation, scale is an exact geometrical operation)

then you could mesh this volume semi-structuredly.
Mesh->Semistructured->Volumes
select the volume (the ring) specifying the amount of divisions in the structured direction (1 in your case)

Mesh->Element type->Prism
and select the volume (by default volume elements are tetrahedra)

GiD will try to automatic detect the semi-structured direction (and select a master-surface), in case of doubt or ambiguity you can splicitly set the direction with
Mesh->Semistructured->Set->Structured direction
or alternativally set the master surfaces of semi-structured volumes
Mesh->Semistructured->Set->Master surface

semi_structured_mesh.png
semi_structured_mesh.png (43.27 KiB) Viewed 4449 times


GiD could also create boundary layer meshes, without require any geometrical modification,
Mesh->Boundary layer
This allow create very thin elements close to some boundaries (but they will be tetrahedra also, no prisms), setting the amount of layers and its sizes following some law.

Re: prism+tets

Posted: Tue Nov 03, 2015 1:25 pm
by ewpostek
Dear Colleagues,
Thank your for your answers.
When writing in the command line
Mescape Geometry Create Reconstruct OneSurfForEachElement
I obtain a message - Reconstruct - unknown command (rather surprising).

This is the attachment with the geo file
as "abelcs" asked to send.

regards,
Elek

Re: prism+tets

Posted: Tue Nov 03, 2015 2:27 pm
by escolano
I attach your model, reseting all your mesh data and settign the ring volume as semi-structured with tetras
layer_2.zip
(206.61 KiB) Downloaded 180 times


About the message "Reconstruct - unknown command" using
Mescape Geometry Create Reconstruct OneSurfForEachElement
probably is because your GiD version (see Help->About) is not modern enough and the command doesn't exists
It exists in last developer versions (current is the 12.1.7d)

Re: prism+tets

Posted: Wed Nov 04, 2015 1:15 pm
by ewpostek
Hello,
Thank you for the model. It helps a lot.

Let me ask:
Is there a way to generate a very thin layer of prism
(they are shb-shell like elements, 6 nodes)?

Supposing, the radius of a sphere is 0.5m,
and the thickness of the layer is 3.0e-04m.
The layer covers the sphere that is
discretized with unstructured tetrahedra,
and the prisms match the discretization of the sphere.

regards,
Elek

Re: prism+tets

Posted: Wed Nov 04, 2015 2:55 pm
by escolano
It doesn't matter if the tickness of your ring is small, it it the same case that I explained.
e.g
create a sphere r=0.5 and create the ring volumes with a copy of surfaces with scale=1.0006 (1.0006=0.5+3e.4/0.5) extruding volumes.
(simple be careful to not to set the 'collapse' checkbutton of the copy window, to avoid join again the close entities)
and set the ring volumes to be meshed semi-structured with 1 division and to be meshed with prisms.

Re: prism+tets

Posted: Thu Nov 05, 2015 8:17 am
by abelcs
Hello,

maybe another option you have is to use the boundary layer mesh (see reference manual). This allows you to define some layers of elements with a given size normal to the surface, avoiding the need to edit the geometry. The thing is that in that case only tetrahedra are generated (not prisms).

Re: prism+tets

Posted: Thu Nov 05, 2015 8:44 am
by ewpostek
Hello,
It is a very good idea for boundary layer in fluids.
However, in this case, the boundary is covered
by the SHB-shell-like elements - 6-nodes.
They are very thin prisms.
Regards, Elek

Re: prism+tets

Posted: Thu Nov 19, 2015 3:17 pm
by ewpostek
ear Colleagues,

I have a surface mesh that I generated with Patran.
I transferred it from patran neutral file
to gid.msh using strings. It means that
the coordinates are exactly the same
as in the neutral file.

I would like to have a volume
(not necessarily sphere) filled
with linear tets.
Therefore, I use
Mescape Geometry Create Reconstruct OneSurfForEachElement

then I attempt to create volume.

I use Create Volume By contour from the menu,
and I select all surfaces.
However, I get the following response;

line number 2 is 3 times into volume generation, must be even
line number 3 is 3 times into volume generation, must be even
line number 6 is 3 times into volume generation, must be even
line number 9 is 3 times into volume generation, must be even
line number 10 is 3 times into volume generation, must be even
line number 11 is 3 times into volume generation, must be even
line number 15 is 3 times into volume generation, must be even
line number 18 is 3 times into volume generation, must be even
line number 19 is 3 times into volume generation, must be even
line number 20 is 3 times into volume generation, must be even
line number 24 is 3 times into volume generation, must be even
line number 27 is 3 times into volume generation, must be even
surface 1 is over 2 surfaces
surface 2 is over 2 surfaces
surface 4 is over 2 surfaces
surface 7 is over 2 surfaces
surface 9 is over 2 surfaces
surface 10 is over 2 surfaces
surface 97 is over 2 surfaces
surface 98 is over 2 surfaces
surface 100 is over 2 surfaces
surface 103 is over 2 surfaces
surface 105 is over 2 surfaces
surface 106 is over 2 surfaces
Problem with lines: 2 3 6 9 10 11 15 18 19 20 24 27
Problem with surfaces: 1 2 4 7 9 10 97 98 100 103 105 106


I do not see anything special in the entities,
however, there is sth. wrong.


Would you mind to give a hint ?


regards,
Elek