add new function to fem module

[bernd]

wow cool! We should definately add this to the examples.

[RatonLaveur]

Regarding the buckling case shown last, might I suggest to lock the end supporting the force in X&Y, this way instead of a full "free end" deflection we see actual buckling?

[twangrt]

wow cool! We should definately add this to the examples.

Ok, then I will add this as well.

I found out what was causing the small difference in results. In the verification example the load are divided equal between all the top nodes, not regarding to the influenced area of the load.

Calculix verification example beam8b.inp

Code: Select all

*NSET,NSET=LAST
    65,  
    66,  
    67,  
    68,  
   101,  
   102,  
   135,  
   136,  
   169,  
   170,  
   203,  
   204,  
   221,  
   238,  
   255,  
   288,  
   289,  
   306,  
   323,  
   340,  
   373,  
   374,  
   391,  
   408,  
   425,  
*STEP
*BUCKLE
10,0.01
*CLOAD
LAST,3,-0.84

FreeCAD generated beam8b.inp

Code: Select all

***********************************************************
** constraints force node loads
** written by write_constraints_force function
*CLOAD
** ConstraintForce
** node loads on shape: Box:Face6
65,3,-3.2812500000000E-01
66,3,-6.5625000000000E-01
67,3,-6.5625000000000E-01
68,3,-1.3125000000000E+00
101,3,-6.5625000000000E-01
102,3,-1.3125000000000E+00
135,3,-6.5625000000000E-01
136,3,-1.3125000000000E+00
169,3,-3.2812500000000E-01
170,3,-6.5625000000000E-01
203,3,-6.5625000000000E-01
204,3,-1.3125000000000E+00
221,3,-1.3125000000000E+00
238,3,-1.3125000000000E+00
255,3,-6.5625000000000E-01
288,3,-6.5625000000000E-01
289,3,-1.3125000000000E+00
306,3,-1.3125000000000E+00
323,3,-1.3125000000000E+00
340,3,-6.5625000000000E-01
373,3,-3.2812500000000E-01
374,3,-6.5625000000000E-01
391,3,-6.5625000000000E-01
408,3,-6.5625000000000E-01
425,3,-3.2812500000000E-01
[code]

[twangrt]

Regarding the buckling case shown last, might I suggest to lock the end supporting the force in X&Y, this way instead of a full "free end" deflection we see actual buckling?

The main goal with that example is to verify that the input file that are generated from FreeCAD match with the verification example from Calculix.

Boring, I know. But still it qualify as actual buckling

[twangrt]

I'm trying to semi-grasp the unit test without reading all the code. Can someone explain to me how it works?

I'm looking at this: https://github.com/FreeCAD/FreeCAD/comm ... 77ec8afd06

It compares the inp file when compiling?

[bernd]

I'm trying to semi-grasp the unit test without reading all the code. Can someone explain to me how it works?

I'm looking at this: https://github.com/FreeCAD/FreeCAD/comm ... 77ec8afd06

It compares the inp file when compiling?

no it does not. This file was mady by me to very easy compile FreeCAD and run a specific unit test manually because a complete unit test run of FreeCAD could take some minutes. It is not used at all from FreeCAD itself. If I need to compile FreeCAD and run a spcific unit test or unit test class I only need to copy the command from this file into my bash and wait a few seconds.

[bernd]

But you are correct. The unit test wors the way you foud yourself. The inp file written by the unit test is compared with a inp file saved in unit test case data directory for calculix test cases. See https://github.com/FreeCAD/FreeCAD/tree ... a/calculix

[bernd]

BTW: The file test_commands.sh is created by Python because I am lazy. Sudhanshu did add the lines manually since he did not know how to create the file by Python. If only one test is added this might even be faster than update the file by Python. https://github.com/FreeCAD/FreeCAD/blob ... ls.py#L103

[twangrt]

BTW: The file test_commands.sh is created by Python because I am lazy. Sudhanshu did add the lines manually since he did not know how to create the file by Python. If only one test is added this might even be faster than update the file by Python. https://github.com/FreeCAD/FreeCAD/blob ... ls.py#L103

I added it manually. Seemed like the easiest way for me.

I'm having second thoughts about the LTB example. I think the example is good, but it would be great to find another verification example, maybe it don't need to be an example from Calculix?

[twangrt]

Now I have updated my lateral torsional buckling example. Hopefully fixed all the CMake/compiling stuff that went wrong in my earlier commits.

To simplify everything the example builds on a verification example from another FEA program.
Example 3.3 in: http://download.strusoft.com/FEM-Design ... amples.pdf

With the current mesh it's within 5% of the example. With more refined mesh(double the current size of the meshpy file) I'm able to get under 3% difference. To note, it that the mesh already is quite big for an example file.

Bernd, have you had time to check anything? I've been struggling a bit with the tests/compiling but I think I've almost got it

Im thinking of trying to fix the name of the returned result files(importCcxFrdResults.py). When calculating with 'buckling' the result is now displayed as 'Time'.

But I have not figured out a specific way to tell that a '.frd' is a buckling analysis result file. My now best bet is to read in the '.dat' file for determining the resultfile.