Probabilistic first ply failure prediction of composite laminates using a multi-scale M-SaF and Bayesian inference approach

Abstract

This paper presents a probabilistic first ply failure analysis of composite laminates using a high-fidelity multi-scale approach called M-SaF (Micromechanics-based approach for Static Failure). To this end, square and hexagonal representative unit cells of composites are developed to calculate constituent stresses with the help of a bridging matrix between macro and micro stresses referred to as the stress amplification factor matrix. Separate failure criteria are applied to each of the constituents (fiber, matrix, and interface) in order to calculate the damage state. The successful implementation of M-SaF requires strength properties of the constituents which are the most difficult and expensive to characterize experimentally, limiting the use of M-SaF in the early design stages of a structure. This obstacle is overcome by integrating a Bayesian inference approach with M-SaF. An academic sample problem of a cantilever beam is used to first demonstrate the calibration procedure. Bayesian inference calibrates the M-SaF first ply failure model parameters as posterior distributions from the prior probability density functions drawn from lamina test data. The posterior statistics were then used to calculate probabilistic first ply failure for a range of different laminates.