Code_Aster fails on blended corner
- Rafael
- Topic Author
- Offline
- New Member
-
Less
More
- Posts: 4
- Thank you received: 0
16 years 4 weeks ago #2803
by Rafael
Replied by Rafael on topic Re:Code_Aster fails on blended corner
Indeed "Netgen 1D-2D-3D" (or "Tetrahedron (Netgen)" + "Netgen 1D-2D") can generate solvable meshes in this case. Thanks for this hint. I must have missed this option in my previous trials.
There is one drawback though. Unfortunately both these algorithms are incompatible with any submeshes that I want to use in the area of big stress concentration. Therefore I have to apply refined mesh to entire model, that leads to very long computation time (or "out of memory" error) on more complicated models. Even though these algorithms generate nicely refined meshes locally on all radiuses, holes and other small features (that is really handy for initial study and I do appreciate this possibility), I'd like to have more direct control over local mesh densities in particular areas in further study.
I'm going to investigate some more possibilities of local mesh refinement on these algorithms (model partitioning, or use Homard and the "MACR_ADAP_MAIL" command, however *.comm files syntax is still black magic for me).
There is one drawback though. Unfortunately both these algorithms are incompatible with any submeshes that I want to use in the area of big stress concentration. Therefore I have to apply refined mesh to entire model, that leads to very long computation time (or "out of memory" error) on more complicated models. Even though these algorithms generate nicely refined meshes locally on all radiuses, holes and other small features (that is really handy for initial study and I do appreciate this possibility), I'd like to have more direct control over local mesh densities in particular areas in further study.
I'm going to investigate some more possibilities of local mesh refinement on these algorithms (model partitioning, or use Homard and the "MACR_ADAP_MAIL" command, however *.comm files syntax is still black magic for me).
- johannes ackva
- Offline
- Premium Member
-
Less
More
- Posts: 157
- Thank you received: 1
16 years 4 weeks ago #2804
by johannes ackva
Replied by johannes ackva on topic Re:Code_Aster fails on blended corner
Rafael, "solvable meshes" can of course be generated by any algorithm. I didn't load Your FE-model into Salome, but since Your .comm-file was good with Kees FE-mesh, Your problem must be in Your mesh. Possibly there are parts where coincident nodes are not merged. Try in the mesh-module the nodes-merging-command. You can even look for not-merged nodes and edges with the help of some of the control-commands in the mesh-module. Non-merged nodes cause the same error "Matrice non factorisable.." as too less boundary conditions
Good luck, Johannes
Good luck, Johannes
- kwou
-
- Offline
- Moderator
-
16 years 4 weeks ago #2805
by kwou
Interest: structural mechanics, solar energy (picture at 'my location' shows too little pv panels)
--
kind regards - kees
Replied by kwou on topic Re:Code_Aster fails on blended corner
Hoi Rafael, Johannes,
Looking at the original mesh of Rafael gives 4 coinciding nodes: 11, 75(!), 76 and 77 (all have same coordinates [x, y, z] = [10, 10, 0] at the corner of the cube).
So indeed coinciding nodes as Johannes stated.
Merge the four nodes in the mesh module by:
kind regards - kees<br /><br />Post edited by: Kees Wouters, at: 2009/04/14 22:42
Looking at the original mesh of Rafael gives 4 coinciding nodes: 11, 75(!), 76 and 77 (all have same coordinates [x, y, z] = [10, 10, 0] at the corner of the cube).
So indeed coinciding nodes as Johannes stated.
Merge the four nodes in the mesh module by:
-
Modification --> Transformation --> Merge Nodes,
default tolerance 1e-5, then Detect
shows nodes 11, 75, 76, 77, (keep this selection)
then Apply
kind regards - kees<br /><br />Post edited by: Kees Wouters, at: 2009/04/14 22:42
Interest: structural mechanics, solar energy (picture at 'my location' shows too little pv panels)
--
kind regards - kees
- Rafael
- Topic Author
- Offline
- New Member
-
Less
More
- Posts: 4
- Thank you received: 0
16 years 3 weeks ago #2810
by Rafael
Replied by Rafael on topic Re:Code_Aster fails on blended corner
Hi there!
you are right. The problem root cause were multiple coincident nodes. Merging them solved issue. I wouldn't figure it out without you guys. Thanks a lot for help!!! Every day I learn something new.
Now I can build very complex "solvable" meshes. I noticed that the easiest workflow is like this:
1) Create linear mesh with all necessary submeshes.
2) Merge coincident nodes.
3) Convert linear mesh into quadratic (in order to increase FEA calculations accuracy).
If in step 1) I create quadratic mesh straight away, usually I get weird distortions in some points (gaps in mesh, partially overlapping triangles etc...)
Best Regards
R.<br /><br />Post edited by: Rafael, at: 2009/04/14 23:02
you are right. The problem root cause were multiple coincident nodes. Merging them solved issue. I wouldn't figure it out without you guys. Thanks a lot for help!!! Every day I learn something new.
Now I can build very complex "solvable" meshes. I noticed that the easiest workflow is like this:
1) Create linear mesh with all necessary submeshes.
2) Merge coincident nodes.
3) Convert linear mesh into quadratic (in order to increase FEA calculations accuracy).
If in step 1) I create quadratic mesh straight away, usually I get weird distortions in some points (gaps in mesh, partially overlapping triangles etc...)
Best Regards
R.<br /><br />Post edited by: Rafael, at: 2009/04/14 23:02
Moderators: catux
Time to create page: 0.146 seconds