#1 2011-05-04 00:01:18

kathi

Member

Registered: 2009-09-24

Posts: 191

SOLVED convergence easier with more layer in an elastoplastic plate?

hello,
I m calculating a plate model (DKT) with elastoplastic material (VMIS_ISOT_TRAC), and also with contact.  When augmenting COQUE_NCOU from 1 to 5 I observed surprisingly that the convergence becomes quite easy. Before (COQUE_NCOU=1) convergence was not reached. I wonder if there is a reason for this? I thought, as more couches, as more heavy work for the solver. Can somebody confirm or esplain me that?

thank you, kathi

cara=AFFE_CARA_ELEM(MODELE=modele,
                    COQUE=(_F(GROUP_MA='MAS0405',
                              EPAIS=0.001,
                              VECTEUR=(0,0,1,),
                              COQUE_NCOU=5 (now) ,  COQUE_NCOU=1 (before)

Last edited by kathi (2011-06-29 09:06:27)

Thomas DE SOZA

Guru

From: EDF

Registered: 2007-11-23

Posts: 3,066

Re: SOLVED convergence easier with more layer in an elastoplastic plate?

Hi,

This is perfectly normal !
When dealing with plates/shells and non-linear computation (and more particularly for material non-linearity) COQUE_NCOU should never be set to 1 since the layers ("couches") are used to integrate over the thickness. With 1 layer (COQUE_NCOU=1), one cannot describe the yielding in the thickness (except of course for a plate that is submitted to uniaxial loading) !
As advised in the U2.02.01, in non-linear analysis COQUE_NCOU should be set to 3 or greater. 5 is a good value.

TdS