Different results in PyANSIS and Classic

[Elslow]

Hello guys!

I'm new in PyAnsys and was trying to test how I could use PyAnsys for a better workflow. I mostly calculating non linear calculations. For this I used a project and run it via PyAnsys mainly with the command mapdl.input() and read in my input files which I'm using in classic batch mode too. So far so go. I get the correct model and the calculation runs fine.
But now I realized that the results are different between PyAnsys and classic. So now the question is if anyone knows why? I want to understand this issue. Is it different due to non linear model or due to the different systems (PyAnsys and classic)?
Just as information on linux classic batch mode I also get different results compared to windows classic. But this I know its due to linux and windows.
The used version is 2023R2 and for PyAnsys I use Phyton 3.11.6
Hope someone can give me a nice answer.

Best Regards,
El_slow

[mikerife]

Hi @Elslow
So one answer is that there should be no difference as PyMAPDL starts and uses an instance of MAPDL. There should also be no difference between running MAPDL on Linux and MS Windows. If there is the MAPDL developers really want to know about that! In my technical support experience differences in results across OSs or hardware is most often due to ill conditioned models being run with too loose of accuracy. I.E. too coarse of a mesh; too large a time/load step; etc.

So here I need to treat this like going to court; I trust what you say but you need to prove it. Can you supply a test model and some comparison of the results (screen shot, spread sheet, whatever)?

What result(s) are you comparing?

Mike

[Elslow]

Hi @mikerife
thank you for your answer. I made a recalculation with a smaller timestep size and now the results are nearly the same in all 3 systems. I thought my stepsize was fine enough but well it isn't.

Samuel

[Elslow]

The Issue was a too large timestep size.