Krande
commented
9 months ago Hey @stefano2734,
I really appreciate you taking the time to provide feedback!
I should probably start off by mentioning that when I created the auto-generated "Ada report" ci release a long time ago I wanted it to do 2 things.
1) Perform a basic comparison of the results from open source solvers Code Aster and Calculix with proprietary solvers that I'm familiar with (Abaqus and Sestra). I just chose an arbitrarily eigenvalue problem for cantilever beam using solid/shell/line elements. 2) Run on every new release of ada-py to ensure that over time the solvers produce similar results (this was primarily to monitor if any of the continuously updated conda-forge dependencies affect the results in any way).
So the idea at the time was not to do any qualitative analysis of the solvers themselves.
However, I do plan on incorporating more qualitative analysis into the auto-generated ADA report in the future.
As of now I still have a few remaining features I want to incorporate into ada-py which allow for better post-processing and visualization of the results.
One of which is compiling a windows version of Code Aster using conda dependencies. Then update the conda-forge feedstock for code aster to include support for windows (windows is my primary work OS) which will help the development process when integrating ada-py with the code aster v16/v17 API's..
With that said, I like the FE benchmarking examples you provided, and I also completely agree with the other points you make about improved description of the problems, references to analytical solution, computational run-time, etc.
So I'll make that a priority once I finish the Code Aster support for windows (which is a task I'll resume in approximately 2 weeks).
I'll just leave this issue open as a reminder to prioritize this task in the next feature release.
stefano2734
Is your feature request related to a problem? Please describe. A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]
here I see not the „problems“ of a solver in quality to theory or best program in the tables.
so is it clear that second order is normally better than first order by same size. Calculation time is significant higher by second order. In complex models fine resolution and first order elements are often better by same solution time.
Describe the solution you'd like A clear and concise description of what you want to happen.
Ada Report: theoretic result values or benchmark nastran in extra column so the reader can see the actual quality of solver in relation to theory or nastran or abacus. Difference in % in extra Table But here the figures of eigenvalues should be the same
Describe alternatives you've considered A clear and concise description of any alternative solutions or features you've considered.
Additional context Add any other context or screenshots about the feature request here.
see other examples of calculix to others: http://www.bconverged.com/benchmarks/index.php
https://www.esrd.com/wp-content/uploads/dlm_uploads/Benchmarks-Guide-Standard-NAFEMS-Benchmarks-Linear-Elastic-Tests.pdf
https://www.autodesk.com/sites/default/files/file_downloads/Autodesk-Inventor-Nastran-2022-Verification-Manual.pdf
https://doc.comsol.com/5.5/doc/com.comsol.help.sme/StructuralMechanicsVerificationExamples.pdf