Only the following list of materials accept a material angle in PART_COMPOSITE. 

21, 22, 23, 33, 34, 36, 40, 41-50, 54-56, 58, 59, 103, 116, 158, 194, 222.

Ushnish 
3/23/10
__________________________________________
lpb on 2009-07-29 at 21:49:18
Revision 54571

Added *PART_COMPOSITE_TSHELL option.
____________________________________________
Reported bug on 10/10/08:  *part_move incompatible with *part_composite.

NLOC in *part_composite has no effect on contact surface when CNTC0 is set to 1 (should
activate shell ref. surface offset in contact).

Note:  Two bugs above were fixed on 10/22/08 and 10/19/08, respectively.

_____________________________________________

If thickness values are given via *element_shell_thickness, those thicknesses become 
the new total shell thickness as opposed to the sum of the integration point thicknesses 
specified with *part_composite.  Each integration point thickness is scaled proportionally.  
Proof:  See the example http://ftp.lstc.com/anonymous/outgoing/jday/shell8.lam22.partcomp.element_shell_thickness.  
This model runs two cases. "case9" uses *element_shell_thickness to set the total shell thickness to 9.0 
whereas "case18" has just *element_shell and the total shell thickness fo 18.0 comes directly 
from *part_composite. Compare the mass in d3hsp, deflection, and integration point stresses in each case.

Something to bear in mind is that shell formulations 2 and 16 are uniform thickness shells and so will use the average of the 4 thickness values given in *element_shell_thickness.  ELFORMs 1 and 6 retain unique nodal thicknesses, i.e., these shells can act as tapered shells.
_____________________________________________
An example utilizing *part_composite is attached (shell8.matmix.partcomp.k) and 
illustrates syntax.  Though 4 different material models are used in this example 
of *part_composite, the material constants are set in each of the 4 material models 
so that each behaves elastically.  This was done to make a direct comparison of 
displacement with an all *mat_elastic benchmark (shell8.all_mat_elastic.k).
This example illustrates that the correct solution is given even when mat_002
(a total Lagrangian material model) is included in the mix. 

The older, but equivalent way of using mixing materials in a single shell 
via *integration_shell is illustrated by shell8.matmix.k

The parameter FAILOPT in *integration_shell is evidently not implemented
for *part_composite.  See ~/test/composite/*failopt*.k.
__________________________________________________________________________
Regarding orientation data in *MAT when multiple mats are used in *part_composite:

The correct way to model this is to use the B1, B2 etc. angles on
*PART_COMPOSITE. We pick up the AOPT and BETA angle from the first layer, or
for the case where the first layer is isotropic, from the first aniosotropic
layer. The other anisotropic layers use the same data as the first. In other words, all anisotropic layers have the same AOPT and BETA values. This is consistent with how we have treated material data in the past.

There was a period of time between the time that I enabled general material
type mixing and the time that I got it straightened out that LS-DYNA was
picking up this data for each layer, but now it is back the way is was.

Best Regards,
Lee Bindeman 3/2009