The tensile strength of unidirectional fiber-reinforced composites is influenced by stochastic fiber breakage and load redistribution. Existing strength models have limitations in predicting fiber break density, clustering, and failure. This study introduces a framework that combines full-field computations based on the fast Fourier transform with a discrete fiber-damage model, yielding a voxelized 3D microstructure in which fibers are grouped into breakage-susceptible blocks according to a Weibull distribution. A stress-based monotonic loading is applied, declaring macroscopic failure when no converged damaged state is achieved. This approach reduces reliance on assumptions regarding load transfer and matrix contributions, facilitating large-scale simulations. It allows for variations in fiber strength distributions and matrix behaviors, independent of initial modeling assumptions. The framework is tested within a community benchmark for consistency with experimental results. After calibrating numerical parameters, it is compared with other models and experiments, showing improved stress–strain predictions, particularly in strength and modulus softening. However, fiber break density remains overestimated, as in existing models. Sensitivity analyses reveal limited influence from fiber debonding length and failure criteria. An alternative set of Weibull parameters aligns stress–strain curves and fiber break density, but may not be valid for single-fiber tests, suggesting caution in extrapolating Weibull laws.
Valmalette et al. (Sun,) studied this question.