LONG_ARC in SSNS501 testcase - please help

k_zurawski

Hi all, I need some help in understanding how to "implement" LONG_ARC in a study which represents SSNS501 testcase. (https://codeaster.gitlab.io/doc/docaster/manuals/man_v/v6/v6.05.501/index.html) I understand how the results are obtained and what arc length method does, but I don't know how should the comm file structure look like. I'm using C_A v15.6
First of all, should I use FONC_MULT to steer load multiplier, or should TYPE_CHARGE='FIXE_PILO' be used insted?
Second thing, I'm using COMPOTREMENT, RELATION ELAS and DEFORMATION GROT_GDEP, should this be commented out and should I use COMP_INCR?
Third question - which node should be the FIXE_PILO, how to decide which one it should be?
Study files:
https://we.tl/t-mP1JMvZXUa

k_zurawski

To answer myself - With FIXE_PILO, you can't use the FONC_MULT keyword for reference loading.
COMP_INCR was replaced by COMPORTEMENT in v12 of C_A.

k_zurawski

[unknown] I don't understand your remark about fascists...
As for the engineering "stuff" - I was dumb enough to forget that I may always import testcase and check how was it done. I used in the mean time ChatGPT with poor results. Answers are misleading. "AI" mixes all possible answers using information from different versions of C_A. For example I don't see any reference node requirement either in the documentation or in the testcase.

As for my patience - well, I don't want someone to look for the answers I already found. Hence I answered myself. I never thought that by writing "to answer myself" can be interpreted as rushing someone. I apologise if anyone felt offended.

Miib, maybe you know why in PILOTAGE a NOM_CMP is defined?

k_zurawski

Hi.
I managed to run the simulation but I'm not getting expected results. My model is not a quarter part of the shell from the testcase. It's full shell. In the testcase, simulation is divided in two non-linear, first part - "normal", second includes arc length method. I simplified the study to have one STAT_NON_LINE with arc-length. Could anyone check my comm file and tell me where I made a mistake? Simulation runs till the end (all 50 time steps converge), but the results seem to show only few % of the applied load. Which seems odd. I'm 100% sure that with the load and BC's set up in this study shell should snap back. I've seen it simulated in Femap/Nastran.

Please help.

Study files:
https://we.tl/t-42jLCoTTPE

jeanpierreaubry

k_zurawski
hello
are you sure of the values for what looks like S235 steel ?
E=3000 seems odd to me

k_zurawski

jeanpierreaubry no, no it was some sort of plastic. Despite the name value is ok. Should be 3GPa, model is in mm.

k_zurawski

I don't understand if PILOTAGE should kick in only in case there's a problem with convergence and before that a normal non-linear simulation should be conducted (as in original comm of the testcase). Or is it possible to use from the beginning of the simulation? When I run the simulation without PILOTAGE I only got to 37% of total load.

AsterO'dactyle

Hello,

You can use PILOTAGE from beginning except for some kinds of PILOTAGE as "DEFORMATION" or "ELASTIQUE".

In addition, selecting the solution of PÏLOTAGE equation (i.e. a quadratic problem with two solutions) with ANGL_INCR_DEPL is problematic because it needs two time steps to work.
You can try NORM_INCR_DEPL or RESIDU.

Your problem is a bit ill-defined.
It is not mathematically possible to use nodal forces for plate/shell elements. In the variational formulation, only linear and surface loads are acceptable.
This is a variational crime.
It should not help convergence (pinching effect).

k_zurawski

AsterO'dactyle It is not mathematically possible to use nodal forces for plate/shell elements. In the variational formulation, only linear and surface loads are acceptable.

I don't understand. In original SSNS501b there's a nodal force applied to COQUE_3D (I attached testcase's comm).

ssns501b.txt

11kB

AsterO'dactyle

Hello,

Unstable non-linear problems like the ones you're dealing with are very difficult to converge in general.

I didn't understand whether you were trying to converge the original test case ssns501b (which must converge!), or whether it was another test ?

k_zurawski

AsterO'dactyle
It was another test. In original testcase there's a quarter part of full model. I have a full model with "hinge" supports on both straight edges of the shell. I'll try to run the original testcase and see if I'm getting any results (which I think I will).

The reason I done it a little bit differently was to learn the process from scratch.

AsterO'dactyle

I've just seen that the support explaining “PILOTAGE” is no longer on the site (training tab). I'm going to look for it

k_zurawski

AsterO'dactyle
I'd appreciate if you could share some info. I read the documentation, some of the coefficients are still unknown for me. I also can't figure out how to limit the angle of the search and how to increase radius. There are few things I understand, but I don't know if the coefficients can be mixed in study (for example ETA_PILO_MAX and ETA_PILO_R_MAX).

k_zurawski

AsterO'dactyle
Hello,
any chance to get the training files on arc-length?

k_zurawski

I ran the simulation again. What I'm getting looks like every second increment applied load is 0 and every second increment is a portion of what should be applied. In other words it looks like loading and unloading every increment. Changing ammount of increments or iterations changes nothing. I imagine that load is running along elastic portion of stability path and reverts followingvthe same path. I'll post a video here later.

k_zurawski

I'm sharing the video I mentioned. It seems that the load is partially applied every second step and removed just after that during next step.
https://youtu.be/n8gXgm-UfwM

AsterO'dactyle

I've put all the slides here:

https://gitlab.com/abbasmic/cours/-/tree/main/codeaster/slides/training/obselete?ref_type=heads

k_zurawski

AsterO'dactyle
Thank you. I see the one regarding arc-length is referring to ssns501 studycase which I want to replicate. Could you look at the video I shared on YT and tell me what's your idea what might be happening?