How to save all the steps up to a point when simulation fails to converge?

k_zurawski

Hi all,
I'm running a nonlinear simulation with elasto-perfectly plastic material. Simulation fails around 92% of the total load. I assume that "structure" loses its stability and applying the rest of the load is not possible. I'd like to follow up the simulation outcomes in paravis up to the point of failure. What I was trying to do is writing base to disk by adding the below line to export file:
R base Example03_3D DRC 0
It worked before with other studies, but those actually converged. After running the analysis I'd like to run POURSUITE to postprocess, but without any outcomes written to disk is obviously impossible. I searched the forum, but either I don't understand what to do or someone employed Python to do the trick (which is black magic for me at this point) Could anyone help me?

example03-3d.txt

4kB

example03-3d.txt

1kB

example03-3d.txt

156kB

lIlIl

jacob Also here: https://forum.code-aster.org/public/d/27380-solved-get-output-for-non-converging-step

k_zurawski

Simulation fails around 92% of the total load

The first thing to try would have been to apply 90% of the final load and inspect the results to see if something unexpected was about to happen.

However, you say it worked with other studies and the message file tells you that it failed to converge after 10 steps. The ConvergenceError message says that you should try again with a larger ITER_GLOB_MAXI. In your case this may work because the RESIDU increased then decreased, so it may converge if you let it do a few more Newton iterations (CONVERGENCE=_F( ITER_GLOB_MAXI=20)).
There are no negative effects to increasing this parameter other than longer computation time on difficult steps (which is what you want).

jacob

many times discussed here...

https://forum.code-aster.org/public/d/27136-stop-properly-running-computation/5

try:
STAT_NON_LINE(...)
except:
pass

k_zurawski

jacob treat me as an ignorant idiot when it comes to programming. I won't feel offended. The only thing I'm able to do is printing "hello world". That's it. I don't understand when where and how, it's not my area of expertise. Maybe someday this will have to change, but now it's not the best time.

k_zurawski

lIlIl I increased ITER_GLOB_MAXI=20 it improved a little bit reaching a little bit higher than before. I must have made a mistake somewhere else, or the structure is supposed to fail at 91% of the load. I'll try with python lines and see if this gives me any result.

jeanpierreaubry

k_zurawski
"everything comes to he who waits"
however learning and practicing a bit of python wont do any harm

k_zurawski

jeanpierreaubry I agree and this is on my list, but not as a priority at the moment. So I imagine that I need to add to the comm file lines (?):
try: Res=STAT_NON_LINE(with rest of my code) except: pass FIN()
Please answer the question, simulation is running about an hour and I don't want to run it again just to know that I did something wrong again.

mf

k_zurawski
Hello,
if you are absolutely not familiar with Python: Python needs proper whitespace, that is 4 spaces before the Res=... and the pass, like if I substitute a space (which is removed by the forum) with a '_':

try:
____Res=STAT_NON_LINE(with rest of my code)
except:
____pass
FIN()

Mario.

k_zurawski

mf that's why I had an error... 🙂

mf

k_zurawski
Guido van Rossum didn't like brackets, so he had to use spaces. So, there's that. 😂

Mario.

k_zurawski

I got the base stored on my disk.
Thank you all!

k_zurawski

One more thing, what's the way to add [SOLVED] comment to the subject of this thread? I don't have a clue how to do it and first thread is locked for me.