PSI - Issue 79
Lorenzo Leonetti et al. / Procedia Structural Integrity 79 (2026) 485–492
488
)
(
1
d
:
: d
,
(6)
t
=
+
coh
RVE
coh
RVE
where coh t and are the cohesive traction and separation rate vectors, respectively. Given the periodic nature of the staggered composite, its nonlinear homogenized constitutive law is derived by considering a suitable Repeating Unit Cell (RUC) subjected to different macrostrain paths under periodic conditions on its boundaries. Finally, in the spirit of coupled-volume homogenization approach (Gitman et al., 2008), the macroscale finite element and the RUC have the same size. As a desirable consequence of this approach, an objective response is found also in the occurrence of strain localization. In a certain sense, this approach produces results which are equivalent to those arising from nonlocal constitutive models based on the injection of regularizing internal length parameters (Tocci Monaco et al., 2021). 3. Proposed data-driven homogenization approach for anisotropic damage evolution This study proposes a novel data-driven homogenization approach for the efficient prediction of anisotropic damage evolution in periodic composites. The proposed strategy is applied in two operational phases, i.e., off-line and on-line phase. In the off-line phase, illustrated in Fig. 2, micromechanical computations are performed on the RUC to derive a complete anisotropic damage evolution law at the macroscale, being able to capture all the potential microscopic failure modes. In the on-line phase, sketched in Fig. 3, the resulting anisotropic damage model is employed for the macroscale failure analyses of composite structures, enabling accurate prediction of multiscale damage mechanisms. The data-driven damage model is used within the coupled-volume homogenization scheme to achieve macroscale mesh-independent results also in the softening regime.
Fig. 2. Flowchart of the proposed data-driven homogenization approach for damaging anisotropic microstructures: off-line phase.
Made with FlippingBook - Online catalogs