Failure Prediction of Fiber Reinforced Composites Using Reduced Order Multiscale Models

Abstract

Fiber reinforced polymer (FRP) composites present a significant opportunity for increas- ing performance and energy efficiency in a number of technology sectors, most notably the automotive and aerospace industries. In order to reduce the development costs for FRP ma- terials, accurate and efficient predictive methods are required which capture the evolution of damage at the heterogeneous microscale. The goal of this dissertation is to advance the state of the art in the failure prediction of FRP composites through new multiscale methods both for the mechanical response and propagation of uncertainty in the material. The contin- ued development of the eigen-deformation based homogenization method with reduced order models (EHM) is presented, including a new approach to address the tension-compression stiffness anisotropy in the fiber direction and a novel parameter weighting method to capture the disparate damage evolution under uniaxial and shear loading. A blind prediction study of laminated IM7/977-3 composites using the improved EHM approach is presented for three composite layups ([0,45,90,-45]2S, [30,60,90,-60,-30]2S, and [60,0,-60]3S) under static tension and compression and tension-tension fatigue with open hole and unnotched configurations. Additionally, Bayesian parameter calibration is implemented within the EHM framework to quantify uncertainty in the composite and is utilized to predict the probabilistic behavior of laminated composite specimens subject to strain rate-dependent effects.