Input Geometries

Submitted by shahabgol on Sun, 12/22/2013 - 07:41

Dear Friends,

I have a few questions about valid input geometries for LIGGGHTS.
What kind of geometries can LIGGGHTS handle? What softwares can create these kind of files?
I have Gambit, Maya, Autocad on my pc. Are there any tutorials for building geometries for LIGGGHTS?
Should I mesh the geometry and export it to LIGGGHTS?

Regards,
Shahab

uhelfenstein | Mon, 12/23/2013 - 09:35

Hi Shahab,

You have to convert your geometry into the STL Format. The STL file as to be in the ASCII Format and not as a binary file. You can import the geometry file for instance with the fix mesh/surface.

Tutorials you can find here:
http://cfdem.dcs-computing.com/?q=node/1689

Best regards,
Urs

shahabgol | Tue, 12/24/2013 - 09:11

Should I mesh the geometry in my geometry maker software or there is no need to export mesh to LIGGGHTS?

Best Reagrds,
SHG

alice's picture

alice | Tue, 12/24/2013 - 12:35

Hello Shahabgol,
well, as indicated by Urs you will need an STL-surface mesh. (But no volume mesh!)
Cheers,
Alice

shahabgol | Tue, 12/24/2013 - 14:24

I have faced this error in my terminal while running my own case. How can I attach my Autocad and STL files to this comment?? or may I mail my files to you so you can help me with my mistakes??
By the way I have compared my own file with the one which I downloaded from the tutorials section and I have noticed the tutorial STL file has an extra mesh on its top. Is it necessary? how can I mesh the top of my cone with autocad??

ERROR:
ERROR: Mesh (id cad1): Mesh elements have been lost / left the domain. Please use 'boundary m m m' or scale/translate/rotate the mesh or change its dynamics (multi_node_mesh_parallel_I.h:548)

cstoltz | Tue, 12/24/2013 - 18:25

Depending on your CAD tool, you can export directly to an STL format (keep it ASCII, not binary), and LIGGGHTS should read it fine. Otherwise, you can mesh it w/ gmsh or similar. Note that if you save to STL directly from the CAD program, you will likely get a minimal number of facets, with the potential for having facets that have very large aspect ratios. This will not parallelize very well, though to date, I haven't noticed much impact on my simulations (except when I calculate stresses on surfaces).

You do need to ensure that the geometry fits inside your simulation cell, or you'll get the error you mentioned. Take note of scaling as it is easy to accidentally save geometry in a different units system than your simulation and have the discrepancy trip you up.

Regards,
Chris

shahabgol | Sun, 12/29/2013 - 08:42

I do not get your last sentence. How can I regulate the scaling in LIGGGHTS? Wh I am amateur in using LIGGGHTS.
What do you mean by "the geometry fits inside your simulation cell"? How can I check this?

Regards,
Shahab

cstoltz | Mon, 12/30/2013 - 18:56

Shahab,

What I mean is that all geometry pieces need to fit entirely within the domain, which was declared near the beginning of your input deck via the create_box command. If your simulation cell is -1

Misterblobby | Fri, 12/27/2013 - 15:11

Hi,

I have had good results by creating the geometries as parts in Autodesk Inventor. Then export each part individually as an ASCII stl file. This keeps the position of all the objects relative to each other in the correct place.

Autocad will create the stl files successfully, but will not reference them to a world coordinate system properly.

shahabgol | Mon, 12/30/2013 - 14:29

What does this error mean?
What should I do to debug it?

ERROR: Mesh cad1: The minumum angle of mesh element 22 (line 156) is lower than the specified curvature. Increase mesh quality or decrease curvature (currently 0.256235°)
ERROR on proc 0: Fix mesh: Bad mesh, cannot continue. You can try setting 'curvature' to 1e-5 or lower (surface_mesh_I.h:475)

Regards,
Shahab

cstoltz | Mon, 12/30/2013 - 19:01

This means that you have a facet in your mesh that has a very high aspect ratio. You can resolve this by either remeshing to get a better quality mesh, or by decreasing the curvature limit via adding the keyword 'curvature' with options to your fix_mesh/surface statement. See documentation for more details on use of 'curvature'.

Regards,
Chris

shahabgol | Mon, 12/30/2013 - 15:05

How about these?

Mesh cad1: element id 0 (line 2) has 6 neighs, but only 5 expected
Mesh cad1: element id 2 (line 16) has 6 neighs, but only 5 expected
Mesh cad1: element id 3 (line 23) has 6 neighs, but only 5 expected
Mesh cad1: element id 4 (line 30) has 6 neighs, but only 5 expected
Mesh cad1: element id 5 (line 37) has 6 neighs, but only 5 expected
Mesh cad1: element id 6 (line 44) has 6 neighs, but only 5 expected
Mesh cad1: element id 7 (line 51) has 6 neighs, but only 5 expected
Mesh cad1: element id 8 (line 58) has 6 neighs, but only 5 expected
Mesh cad1: element id 9 (line 65) has 6 neighs, but only 5 expected
Mesh cad1: element id 10 (line 72) has 6 neighs, but only 5 expected
Mesh cad1: element id 11 (line 79) has 6 neighs, but only 5 expected
Mesh cad1: element id 12 (line 86) has 6 neighs, but only 5 expected

ERROR: Mesh cad1: elements 0 and 147 (lines 2 and 1031) are duplicate
ERROR on proc 0: Fix mesh: Bad mesh, cannot continue. You can try re-running with 'heal auto_remove_duplicates' (surface_mesh_I.h:456)

cstoltz | Mon, 12/30/2013 - 19:03

This means you have a bad mesh and need to remesh it with a different utility. As the message says, you can try adding the 'heal_auto_remove_duplicates' keyword, but make sure you check the resulting geometry that it generates to make sure nothing important was lost.

A better option in my opinion would be to go back to the software you used to generate the mesh and try to redo it better. Another alternative is to use a software like MeshLab that can manipulate STL's well, and use it to remove duplicates, zero area facets, etc.

Regards,
Chris