PSO-driven micromechanical identification of in-situ properties of fiber-reinforced composites

Abstract Within the framework of micromechanics, the newly expanded Generalized Finite-Volume Direct Averaging Micromechanics is connected to the gradient-free Particle Swarm Optimization (PSO-GFVDAM) to identify the in-situ constituent properties and residual stresses and study their effects on the effective and localized responses of fibrous composites. The present technique is advantageous by avoiding the gradient updating concept and adopting the advanced generalized finite-volume micromechanics with plasticity effects, both of which guarantee the stability and efficiency of the proposed technique. After introducing the microstructural effects from the fabrication process and chemical reaction that cannot be easily detected from the macroscopic behavior of composite materials, the PSO-GFVDAM is employed to identify the elastic constituent properties of carbon/epoxy composites, yield stress and hardening parameters of the aluminum-matrix system, as well as the residual stress state within the microstructures. The stability and accuracy of the algorithm are tested by checking the iterated particle distributions, errors generated at each step and substituting the deduced parameters back into direct analyses to predict effective responses of off-axis loaded specimens. More important to the effective response, the localized stress distributions are efficiently recovered, helping to characterize the possible damage initiation and crack propagation that usually start from the material levels.