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

You are not logged in.

#1 2022-05-13 10:24:58

Hansbau52
Member
From: Germany
Registered: 2022-05-10
Posts: 8

SHELL/3D versus SHELL/3D-SI

Hy community,

My test model (see attached picture) with linear elements (3D/DKT) produces a deformation of 1.77 mm. 
There is no difference between 3D and 3D-SI.  The same calculation with Abaqus gives a displacement of 1.79 mm with C3D8/S4 elements. If I replace C3D8 with C3D8R (reduced integration), I obtain a displacement of 2.64 mm.

I expect a result around 2.5 mm ( beam, analytically FL³/3EI). I also obtain this with a pure shell model.
Have I missed something (3D versus 3D-SI)?

Best Regards,

Hansbau52

Last edited by Hansbau52 (2022-05-13 10:36:38)


Attachments:
Screenshot_01.jpg, Size: 59.8 KiB, Downloads: 99

Offline

#2 2022-06-22 10:49:11

Johannes_ACKVA
Member
From: Ingenieurbüro für Mechanik, DE
Registered: 2009-11-04
Posts: 756
Website

Re: SHELL/3D versus SHELL/3D-SI

hello,

you should obtain a very good result wrt beam theory, if you use quadratic elements (HEXA20  instead of HEXA8). And set NU=0 (because a beam has no lateral contraction)

Moreover you must take care how to attach the DKT-elm to the 3D-elm. If you have fused (shared) nodes with an overlapping, then your model is too stiff in this region by factor 2. It is better done without overlapping with the glue contact:
AFFE_CHAR_MECA(..LIAISON_MAIL( .. TYPE_RACCORD='COQUE_MASSIF' . .

Best regards
Johannes_ACKVA

*** NEW: all Code-Aster courses now  as Video and Webinar courses  ***


Ingenieurbüro für Mechanik
D 91717 Wassertrüdingen / Germany

www.code-aster.de                                                Training & Support for NASTRAN and CODE-ASTER

Offline

#3 2022-06-28 12:19:48

Hansbau52
Member
From: Germany
Registered: 2022-05-10
Posts: 8

Re: SHELL/3D versus SHELL/3D-SI

Hello Dr. Ackva,
thank you very much for your answer. I had of course set the cross-contraction to zero and also used the connection with COQUE-MASSIF without overlap. My question was about the reduced integration for the linear 3D elements. In Abaqus it is standard and leads to relatively good results. With Code_aster I did not notice any difference. Can you please tell me something about this?

Best regards,
Hansbau52

Offline

#4 2022-06-28 12:36:55

Johannes_ACKVA
Member
From: Ingenieurbüro für Mechanik, DE
Registered: 2009-11-04
Posts: 756
Website

Re: SHELL/3D versus SHELL/3D-SI

hello

your questions are answered in  doc U2.01.10:

subintegration (modelisation='3D_SI' instead of '3D')  meaningful when dealing with (nearly) incompressible materials,

quadratic elements (HEXA20  instead of HEXA8, TETRA10 instead of TETRA4) recommended for the mechanical analyses

Best regards
Johannes_ACKVA

*** NEW: all Code-Aster courses now  as Video and Webinar courses  ***


Ingenieurbüro für Mechanik
D 91717 Wassertrüdingen / Germany

www.code-aster.de                                                Training & Support for NASTRAN and CODE-ASTER

Offline

#5 2022-06-28 13:47:23

Hansbau52
Member
From: Germany
Registered: 2022-05-10
Posts: 8

Re: SHELL/3D versus SHELL/3D-SI

Hello Dr. Ackva,

of course, the square elements are preferable. I just wanted to understand with the linear elements the difference in approach Abaqus/Code_aster.  As I said, in Abaqus the under-integration is set as default for linear quadrilateral elements (2D) and also for the linear hexahedra. In addition, any control parameters are still used. The intention is to get a grip on locking and hourglass effects. I think we should end the discussion with this and I thank you expressly for your hints.

Best regards,
Hansbau52

Offline