Abstract
A finite element procedure for simulating damage progression in an advanced composite material loaded in fiber-direction tension has been developed. The eight-node, eight-degree-of-freedom element is used to model a small cell of a composite, the properties of which were determined from micromechanical simulations. A Monte Carlo simulation technique is used in the micromechanical simulation to characterize the random degradation paths that this cell can follow when subjected to axial tension. These results are synthesized using a neural network to efficiently determine the expected value and variance of the degradation of the random-damage finite element given the current state of the element. Results from failure simulations of Kevlar/epoxy composites are presented and compared to experimental results.