Modeling of element imperfection

[lukaszlaba]

I have made some test for nonlinear analysis with anastruct.
First AISC 360 benchmark was checked.
https://notebooks.azure.com/lukasz-laba1/projects/ansastruct/html/AISC%20360%20benchmark-20190301.ipynb
All considered cases agree with AISC benchmarks. It's Great !!
Then I made some test with imperfection modeling (big and small delta).
https://notebooks.azure.com/lukasz-laba1/projects/ansastruct/html/DAM-test_20190228.ipynb
The idea was to create and calculate one by one deformated structures and finally get envelope of results.
I think about some new features in anstruct to make analysis like this more easy:

• method to make curved member with small delta parameter is needed,
• parameter for big delta effect is needed,
• some idea for making envelop is needed.

I thinking about anaStruct contribution. Does this idea for direct nonlinear analysis could be right way to anaStruct upgrade?

[ritchie46]

Hi, Really cool that you have made the AISC benchmark. I did include a geometric stiffness matrix, but did not do exhaustive testing yet, so this is great! I'll add these to the tests of the package. It should be noted that the geometic stiffness matrix is an approximation, so the results would be more accurate with small delta's. If you want more accurate result with big delta's we should probably add more elements to the grid. I've based it on this document. http://people.duke.edu/~hpgavin/cee421/frame-finite-def.pdf

I don't really know what you mean by envelope results, could you refer to some documentation about that so I get an idea of what you want/ mean?

[lukaszlaba]

Result envelope mean to have displayed results for many load cases on one diagram.
http://www.bgstructuralengineering.com/BGSMA/ContBeams/BGSMA_CB_04.htm
So when different imperfections are considered in fact we get series of results and it is nice to see them on one diagram to easy get know where the maximum for example of bending moment occur.
My current concept is to collect calculated SystemElements instances and then in some way display those results on one diagram. Probably modified ss.show_.. () methods can be used to do it. So maybe new class named res_bank() is needed instead of including this mechanism inside SystemElements class. Please let me know if there is better way to do it.

[ritchie46]

Ah.. Now I know what you mean. I know the concept, but did not know the english term. ;) I think this is quite feasible to implement for loads in combination with the LoadCombination class. . The numerical result for this can be found by finding the maximum and minimum values of a load combination. However it does need some adaptation for the plotter to get a nice visual result back.

For loads I will take a look on this.

For element inperfections it is a little bit harder. You indeed need multiple SystemElement objects. If I needed it for a project I would stack the numerical results in a ndarray and plot them differently.