|
Finite Element
Technology:
- Solid-Shell
elements: A correct reproduction of severe thickness changes in
thin-walled structures can be accurately described by the use of
three-dimensional solid elements, especially under contact situations.
Compared with shell formulations based on plane-stress assumptions,
natural derivation of constitutive laws is obtained with solid
elements, in addition providing a straightforward extension to
geometrically non-linear problems. Contrary to shell elements, possibly
double-sided contact situations can be accurately considered (due to
the presence of physical nodes on top and bottom surfaces). These
advantages make this class of elements desirable, for instance, in the
simulation of sheet metal forming processes. Among solid elements, the
low order eight-node brick element is widely used given its simple
formulation and robustness. However, the major drawbacks of brick
elements reside in its strong sensitivity to locking phenomena and poor
computational performance due to the use of multiple element layers
when applied to thin walled structures. Especially for modeling shell
structures with bending, as thickness to length ratio tends to zero,
the transverse shear-locking and thickness locking phenomena become
more evident. Also, plasticity (or incompressible elastic materials)
leads to isochoric deformation, which is the main source of the
volumetric locking phenomenon. The Enhanced Assumed Strain Method is
used to tackle efficiently locking pathologies
Constitutive laws for
anisotropy:
Two approaches might be used to account for general
three-dimensional plastic anisotropy:
-
Phenomenological
non-quadratic yield functions combined with an isotropic strain
hardening law. The well-known Barlat’s yield function yld2004-18p
already proven to be accurate and CPU efficient, but require a
substantial number of input parameters
-
Polycrystal
plasticity models. Polycrystal models naturally account for anisotropy
and texture evolution using only initial texture as input. However,
they are known to be computationally heavy. Particularly, the
rate-independent approach proposed by Gambin is utilized. This model
avoids the uniqueness issue on the choice of active slip systems, by
applying a regularized Schmid law. The corresponding yield surfaces
have smooth corners, and the strain-rate normal vector is uniquely
defined.
|
