Interface stress distributions in transversely loaded continuous fiber composites: parallel computation in multi-fiber RVEs using the boundary element method

We present the results of a computational investigation on the effect of microstructure on the (spatial and statistical distribution) of interface stresses in transversely loaded continuous fiber-reinforced composites. Instead of traditional unit-cell models, we use a parallel implementation of the boundary element method (BEM) to analyze large representative volume elements (RVEs) containing up to 144 individual fiber cross-sections placed within the domain of interest using a Monte Carlo (MC) algorithm. The BEM is well suited for this multi-inclusion modeling because of its relative accuracy and the ease of meshing complex geometries. By controlling the minimum allowable inter-fiber spacing, the MC algorithm allows for the generation of microstructures showing various degrees of homogeneity (or, disorder). Spatial correlation functions and overall properties are computed on these multi-fiber RVEs to ensure that their size allows for appropriate statistics to be inferred. After validating the computer code and giving some numerical examples, the effects of key microstructural and material parameters, namely the minimum inter-fiber spacing (δ), the fiber/matrix stiffness ratio (Ef/Em) and the fiber volume fraction (φ) on the stress distributions are examined. Particular attention is placed on the statistics of the distribution of the maximum interface stresses computed on each fiber; these are found to follow a Weibull-like distribution, whose specific shape depends on material and microstructural parameters. This information can be readily incorporated into statistical models for the prediction of composite failure.