Maybe basic questions, user interface and solver/meshing options
205 views
Skip to first unread message
Raúl Arrese
Dec 15, 2014, 12:11:21 PM12/15/14
to agr...@googlegroups.com
Hi,
I am using the soft for some weeks in a steady state heat trasfer application, I have an electrical engineering background without expertise in FEM but with a basic kwnoledge of the method and PDEs.
I would like to post some comments from my experience from user interface and dobuts relating the (for me) complex parameters of the soft that can not understand.
BTW is the best FEM software out there that could grap and understand what was doing almost from scratch. Great work.
User interface:
After doing some simple tests to feel how to define boundary conditions, materials, etc. I imported a DXF file, the unconnected nodes were shown with segments in red and I corrected them quickly.
After that I could not see any red segment or node (like 800 total nodes) but when meshing it ketp saying some nodes not fully connected. It was painstaking finding them as no clue about node id or whatever was given, and were not clearly highlighted on screen. Finally I edited the DXF file, simplified a lot the curve polylines (to 300 nodes and edges) and it seemed to work.
Now it appeared "nodes in a loop" or something like that error message. It took a while to understand what was the meaning, and same as before no clue about what nodes where affected. I exported to DXF and manually cut it in pieces and loaded it again to see where was the "bad loop". Finally got it repaired.
So my suggestion: please tell ID number or highlight the nodes affected by the error messages.
Meshing-Solving.
I don´t know very well how to use or select the meshing options, matrix solver selection, and linear/non linear problem selection.
I.e. if my problem is linear heat transfer steady state (Laplace), why should I select non linear solver?
It would be very useful at least a summary guide to them.
I found memory error issues when I increase the mesh options from the very basic. i.e. I only get a solution with:
- triangle mesh, polinomial order 2 and 0 refinements
- triangle mesh, polinomial order 1 and 1 refinements
with other meshes, i.e. polinomial order 2 and 1 refinement it tries but seems to memory overflow.
I read some basics about HP adaptivity and seems useful but should understand first the parameters before try to use it. Any help out here on that?
As example this is the output for triangle meshing, polinomial order 2 and 1 mesh refinement.
With more detailed meshing as triangle-quad fine, gmsh or similar the mesher gives memory allocating errors also (it does not get even to start the solver).
Start: 17:46:04.195
Comments welcomed, I know I don´t have a thorough understanding of the theory details, but please understand, I am learning this as not-paid-by-boss work as really think that it deserves the effort.
In any case, great work
Regards
Raul
I am using the soft for some weeks in a steady state heat trasfer application, I have an electrical engineering background without expertise in FEM but with a basic kwnoledge of the method and PDEs.
I would like to post some comments from my experience from user interface and dobuts relating the (for me) complex parameters of the soft that can not understand.
BTW is the best FEM software out there that could grap and understand what was doing almost from scratch. Great work.
User interface:
After doing some simple tests to feel how to define boundary conditions, materials, etc. I imported a DXF file, the unconnected nodes were shown with segments in red and I corrected them quickly.
After that I could not see any red segment or node (like 800 total nodes) but when meshing it ketp saying some nodes not fully connected. It was painstaking finding them as no clue about node id or whatever was given, and were not clearly highlighted on screen. Finally I edited the DXF file, simplified a lot the curve polylines (to 300 nodes and edges) and it seemed to work.
Now it appeared "nodes in a loop" or something like that error message. It took a while to understand what was the meaning, and same as before no clue about what nodes where affected. I exported to DXF and manually cut it in pieces and loaded it again to see where was the "bad loop". Finally got it repaired.
So my suggestion: please tell ID number or highlight the nodes affected by the error messages.
Meshing-Solving.
I don´t know very well how to use or select the meshing options, matrix solver selection, and linear/non linear problem selection.
I.e. if my problem is linear heat transfer steady state (Laplace), why should I select non linear solver?
It would be very useful at least a summary guide to them.
I found memory error issues when I increase the mesh options from the very basic. i.e. I only get a solution with:
- triangle mesh, polinomial order 2 and 0 refinements
- triangle mesh, polinomial order 1 and 1 refinements
with other meshes, i.e. polinomial order 2 and 1 refinement it tries but seems to memory overflow.
I read some basics about HP adaptivity and seems useful but should understand first the parameters before try to use it. Any help out here on that?
As example this is the output for triangle meshing, polinomial order 2 and 1 mesh refinement.
With more detailed meshing as triangle-quad fine, gmsh or similar the mesher gives memory allocating errors also (it does not get even to start the solver).
Start: 17:46:04.195
Mesh Generator: Initial mesh generation
Solver (heat): Field solving (single analysis)
Solver: Elapsed time: 00:05.761 s
Mesh View: Initial mesh with 98320 elements
Mesh View: Linearizer (initial mesh) processing failed: Hermes::realloc_with_check() failed to reallocate 132732000 bytes.
Mesh View: Solution mesh with 98320 elements
Mesh View: Linearizer (solution mesh) processing failed: Hermes::realloc_with_check() failed to reallocate 132732000 bytes.
Mesh View: Polynomial order
Post View: Scalar view (heat_temperature_degrees)
Mesh View: Linearizer (scalar view) processing failed: Hermes::realloc_with_check() failed to reallocate 132732000 bytes.
Comments welcomed, I know I don´t have a thorough understanding of the theory details, but please understand, I am learning this as not-paid-by-boss work as really think that it deserves the effort.
In any case, great work
Regards
Raul
Reply all
Reply to author
Forward
0 new messages