ELAS_COQUE is fundamentally flawed

todd_alan_martin · 2010-03-03 07:13:48

Hi

I have been reading the documentation for DEFI_MATERIAU with ELAS_COQUE

The definitions of the C, D and G matrices for an equivalent Isotropic material contain the thickness dimension, h, so the three stiffness matrices do not have identical units, N/m2, as I would expect.

Should the documentation read
C/h=[C1111,C1122.....]
D/h3=[D1111,D1122....]
G/h=[G11,G22]

for consistent units?

Otherwise, how does the solver know the plate thickness to calculate the material stiffness properties in membrane, bending and shear?

Last edited by todd_alan_martin (2010-03-04 06:10:16)

todd_alan_martin · 2010-03-04 06:03:06

In answer to my own question. I have determined that the documentation regarding the definition of a shell material using ELAS_COQUE is unfortunately correct.

Let's say I have a model containing steel plate with thicknesses of 2mm, 5mm and 10mm.
If I used ELAS_ISTR I can define one material "steel" and assign it to the 3 different plate thicknesses using AFFE_CARA_ELEM.

However, if I want to include transverse shear (MODELISATION=DST) using ELAS_COQU, I have to define 3 separate materials, "steel1", "steel2", "steel3", because the plate thickness is included in the material definition.

Has anyone considered how stupid this is? The stiffness matrix should be in standard units of Pascals.

Last edited by todd_alan_martin (2010-03-05 00:16:56)

desroches · 2010-03-08 18:05:48

Hi,

Thank you for your message and your pertinent remark. It is true that the thickness should not be in DEFI_MATERIAU, but instead in AFFE_CARA_ELEM like in other cases.
But as we wanted to introduce C,D and G matrices, and not C/h, D/h3,etc.. (which 'd have no physical meaning), we prefered to do so.
I must add that we use ELAS_COQUE very rarely and hadn't noticed this problem yet. Maybe in another context, we would have changed this.

todd_alan_martin · 2010-03-09 08:23:35

Hi Desroches

Thanks for your reply. Since you use ELAS_COQUE rarely, surely it can be easily changed. Or you could introduce another command DEFI_COQU_HOMO
The formulation I have outlined is used by Nastran and most other, if not all, commercial solvers.

Actually the bending matrix is represented by 12D/h3, so that it is identical to the membrane matrix, C/h for an isotropic material.

Apart from this.
Is it possible to convert DEFI_COQU_MULT into ELAS_COQU, so that a laminate can be treated as a homogeneous plate? It is often convenient to study the stresses and strains at the middle, top and bottom surfaces, rather than having to output data for the separate layers.

Todd.

desroches wrote:

Hi,
Thank you for your message and your pertinent remark. It is true that the thickness should not be in DEFI_MATERIAU, but instead in AFFE_CARA_ELEM like in other cases.
But as we wanted to introduce C,D and G matrices, and not C/h, D/h3,etc.. (which 'd have no physical meaning), we prefered to do so.
I must add that we use ELAS_COQUE very rarely and hadn't noticed this problem yet. Maybe in another context, we would have changed this.

Last edited by todd_alan_martin (2010-03-09 11:19:07)

#1 2010-03-03 07:13:48

ELAS_COQUE is fundamentally flawed

#2 2010-03-04 06:03:06

Re: ELAS_COQUE is fundamentally flawed

#3 2010-03-08 18:05:48

Re: ELAS_COQUE is fundamentally flawed

#4 2010-03-09 08:23:35

Re: ELAS_COQUE is fundamentally flawed

Board footer