Help with composite plate with 28 plies takes a long time to run

alberto · 2017-04-25 19:46:36

I've been testing code aster for its use with composite materials. For small configurations it works quite well. However, I was running a test in a simple rectangular panel subjected to a constant pressure and it is take quite some time. The mesh is created in gmsh and it is not that big. It has 1500 quad4 DSQ elements.
Looking at the log it seems to be taken most of the time in MECA_STATIQUE. I would expect that but from what I gather in the log is not in solving the system but in assembly it:

#1 Resolution des systemes lineaires CPU (USER+SYST/SYST/ELAPS): 0.96 0.02 0.19
#2 Calculs elementaires et assemblages CPU (USER+SYST/SYST/ELAPS): 625.08 0.21 625.40

A laminate with 28 plies is not that estrange when working with thin pre-pregs. This example could be simplified a good deal giving that I'm using just 0/90 plies. However, that is often not the case. 10 minutes in a i7 with plenty of ram is quite a long time for this problem. I'm wondering if I'm doing something wrong in some configuration. I've been trying to change values for maximum memory and so on without success. Is this kind of running time expected in this situation?
I would appreciate some pointers into what can be done to improve this.
I'm running version 13.3 on Ubuntu and I have attached the mess file plus mesh and comm files.

Attachments:
composite_plate.tar.gz, Size: 21.3 KiB, Downloads: 365

alberto · 2017-05-02 08:14:25

I've been trying different things to alleviate this problem. I've found an interesting thread at:
viewtopic.php?id=14683 (sorry but it seems I can not post links)
Since I'm using some python code that takes elemental forces and applies them to the laminate using CLP theory, I do not need stress. However, when I run the case using OPTION='SANS' it just takes the same time, the difference is thus it just spends it at CALC_CHAMP.
Other alternative would be to use an anisotropic shell by inputting matrices directly. I've tried that and I get results that are wildly different.
I've also tried using global strains by calculating with 'SANS' and converting those shell strains to layer strains and stresses (that's what I'm after) but I get results that are not consistent with reference cases and other software.
Does anybody have another idea on how to reduce the computation time?

todd_alan_martin · 2018-09-27 08:02:12

Are you calculating transverse shear results? See topic id=14683

#1 2017-04-25 19:46:36

Help with composite plate with 28 plies takes a long time to run

#2 2017-05-02 08:14:25

Re: Help with composite plate with 28 plies takes a long time to run

#3 2018-09-27 08:02:12

Re: Help with composite plate with 28 plies takes a long time to run

Board footer