In this paper, micromechanical simulations are employed to evaluate the performance of the Tsai–Wu and Hashin failure criteria for fiber-reinforced composites, especially in stress states whose experimental reproduction is complicated. Micromechanical responses are generated using a finite element model of a representative volume element, in which only the matrix material experiences damage and the fibers are assumed to be elastic. Micromechanical simulations of basic load cases are used to calibrate macrolevel criteria. Finally, the response of the micromodel and macromodels is compared for various load combinations. Despite a good agreement between Tsai–Wu criterion predictions and micromodel results in a wide range of stress states, some stress combinations are highlighted for which the strength is not predicted accurately. Additionally, accuracy of the Hashin criterion suffers from ignoring the influence of stress in fiber direction on matrix failure.

Micromechanical simulation of composite material failure requires a pressure-dependent failure model for the polymeric matrix. Available pressure-dependent damage formulations assume a certain shape of the stress-strain law under uniaxial loading. However, upon close inspection none of the available formulations is able to reproduce the assumed shape. This implies that input values for the fracture energy cannot be recovered exactly. In this paper, a new methodology for developing consistent pressure-dependent damage models for polymeric materials is presented. Using this method the predefined shape of the stress-strain relation of an element with localized deformation under uniaxial tension can be exactly reproduced which enables further to recover the exact amount of energy dissipation consistent with the input toughness. The methodology is demonstrated for two different softening laws, namely linear and exponential softening. These models are applied to the damage analysis of unidirectional continuous fiber-reinforced composites. The formulation is validated by simulation of a test for Mode-I fracture energy characterization and comparing the load-displacement response with that obtained with cohesive elements.