Numerical methods for delamination analysis, such as the cohesive zone method, require fracture energy as an essential input. Existing formulations rely on a phenomenological relationship that links fracture energy to the mode of fracture based on linear elastic fracture mechanics (LEFM). However, doubts exist about the applicability of LEFM. It has been demonstrated that the phenomenological relationships describing fracture energy as a function of mode-ratio are not universally valid. Computational homogenization (FE²) provides an alternative where the dissipative mechanisms can be resolved on the microscale. This paper aims to assess the suitability of a proposed discontinuous FE² framework for characterizing delamination growth under mode-II conditions by comparing it to direct numerical simulations (DNS). The impact of plasticity on effective fracture energy is evaluated for two distinct mode-II test configurations. The dissipation density from the bulk integration points within the delamination propagation zone is monitored. The findings demonstrate the FE² model's capability to accurately capture plastic energy dissipation around a growing crack. Variations in plastic dissipation are observed between the mTCT and ENF test setups, leading to differences in effective mode-II fracture energy. These nuances, unaccounted for in state-of-the-art mesoscale cohesive models, highlight the FE² framework's potential for enhancing delamination modeling.

A common choice for multiscale modeling of the mechanical response of composites is to use periodic boundary conditions (PBCs) on square and cubical representative volume elements (RVEs). However, when strain localization occurs in the micromodel, these PBCs are unable to reproduce the transverse isotropy of composite materials with a random microstructure. Existing remedies to alleviate this issue have been proposed in literature by either rotating or shifting the periodicity constraints. However, this results in a mismatch of the microstructure on opposing edges which may prevent cracks to cross the boundary and consequently limit the supported localization angles. Furthermore, in absence of a strategy that ensures a single localization band to arise in a fracturing RVE, it is difficult to formulate a generic expression for the length scale parameter that is used to regulate the energy dissipation, which plays an important role in obtaining RVE-size objective results. As an alternative to square (or cubical) RVEs, circular (or spherical) RVEs have been proposed in literature since they provide a response which is independent of the orientation due to shape of the RVE. However, it is shown in this work that the existing formulation with straightforward application of PBCs on a circular RVE fails to predict the correct softening behavior, due to over-constraining when cracks reach the boundary. Therefore, a new formulation of PBCs on a circular RVE is proposed, which allows for a single fully developed localization band under arbitrary angle. The performance of the new formulation is tested with a series of simulations where macroscopic strains are imposed under varying orientations. It is demonstrated that the circular RVE with the new formulation of PBCs successfully predicts a transversely isotropic response with full softening without the issue of mismatching microstructure as with previously developed remedies for the square RVE. In addition, it is shown that the length scale parameter is well-defined and independent of the orientation of the circular RVE.

The mode-I dynamic fracture energy and failure mechanisms of glass fiber-reinforced polymer composites are investigated with an embedded cell model of the single-edge-notched-tension (SENT) geometry. Under an applied dynamic loading, a crack may propagate in the embedded microstructure, accompanied by the development of a fracture process zone in which fiber/matrix debonding, matrix cracking and ductile matrix tearing are observed. Reaching a maximum nominal strain rate of 250/s, a series of SENT tests are performed for different loading velocities and specimen sizes while the dynamic energy release rate is evaluated using the dynamic version of the J-integral. The influence and interaction of loading rate, time-dependent material nonlinearity, structural inertia and matrix ligament bridging on the fracture toughness and failure mechanisms of composites are evaluated. It is found that with the given material parameters and studied loading rate range, the failure type is brittle with many microcracks but limited plasticity in the fracture process zone and a trend of increasing brittleness for larger strain rates is observed. The inertia effect is evident for larger strain rates but it is not dominating. An R-curve in the average sense is found to be strain-rate independent before the fracture process zone is fully developed and afterwards a velocity–toughness mechanism is dictating the crack growth.