Welcome to the forums. Please post in English or French.

You are not logged in.

#1 2021-01-13 13:25:44

Fabrizio
Member
Registered: 2020-11-20
Posts: 6

Convergence problem in nonlinear analysis

Hello friends, I hope you are well!
I am analyzing a case of irradiation, using the aster's own IRRAD3M law, and after almost an hour of running, almost 300 iterations I have problems in convergence. Can you help me? attached the message that appeared.
Thank you very very much


Attachments:
message, Size: 100.23 KiB, Downloads: 11

Offline

#2 2021-01-14 10:16:57

mf
Member
Registered: 2019-06-18
Posts: 145

Re: Convergence problem in nonlinear analysis

Hello,

I am no expert for irradiation, but the message says your calculation converged, but there was not enough time to complete the following step? Weird, but if you add some time for your calculation it should complete (tpmax or time_limit should be higher).

I do not know exactly the 'rules' if you tell CA RESI_GLOB_RELA and RESI_GLOB_MAXI (one has priority, that is what I mean), but telling it one of those values (I only use RESI_GLOB_RELA for example) is enough.

If you aim for less calculation time with OpenMP (meaning without MPI), you should choose ncpus=number_of_cores/2 (turning off hyperthreading on Intel CPUs also helps). You chose ncpus=1 ('credit of processes used:  1'). So there might be some time to gain.

Hope it helps,

Mario.

Last edited by mf (2021-01-14 10:22:03)

Offline

#3 2021-01-15 01:46:30

Fabrizio
Member
Registered: 2020-11-20
Posts: 6

Re: Convergence problem in nonlinear analysis

Hello dear
How do I do this part ?:
If you aim for less calculation time with OpenMP (meaning without MPI), you should choose ncpus = number_of_cores / 2 (turning off hyperthreading on Intel CPUs also helps). You chose ncpus = 1 ('credit of processes used: 1'). So there might be some time to gain.

Cheers

Offline

#4 2021-01-15 07:53:21

mf
Member
Registered: 2019-06-18
Posts: 145

Re: Convergence problem in nonlinear analysis

Hello,

in Code_Aster: you edit the .export file. If you have, let's say an 8-core CPU you edit it to

P ncpus 4

in Salome-Meca: you change the entry in 'Number of threads', although SM should automatically use the right number of threads (although it only shows in the .mess-file, not in the window 'Run parameters').

Mario.

Last edited by mf (2021-01-15 07:54:39)


Attachments:
Bildschirmfoto vom 2021-01-15 07-52-04.png, Size: 79.24 KiB, Downloads: 10

Offline