Session: 03-08-01: Micromechanics and Multiscale Modeling

Paper Number: 137515

137515 - Protocols for the Estimation of the Extreme Value Distributions of Stresses Driving Damage Initiation in Polymer Matrix Composites 

Polymer matrix composites (PMCs) reinforced with continuous fibers have become structural materials of choice in failure critical aerospace applications, primarily owing to their low density, excellent corrosion resistance, and high ductility. Undesired variable performance of PMC components is known to stem from microstructure stochasticity. Further adoption of PMCs in structural applications is currently hindered by the lack of protocols for efficient exploration of microstructure-damage resilience correlations covering a large microstructure space of interest. To elaborate, the primary challenge in formulating the desired low computational cost (reduced-order) model comes from the ability to generate a diverse set of PMC microstructures and the need to identify the salient features of the generated PMC microstructures. A second challenge comes from the need for numerically robust protocols for calculating predicted failure indicators based on the stress fields predicated by micromechanical finite element simulations. Our work addresses both these challenges.

Specifically, we will present advances in (i) tools used to generate stochastic instantiations of PMC microstructures for varying fiber volume fractions, orientations, levels of clustering, and anisotropy of clustering, (ii) rigorous quantification of PMC microstructures using the 2-point spatial correlations framework and their low-dimensional representation using principal component analyses, and (iii) consistent extraction of parameters describing the extreme-value distribution (EVD) of failure indicators from finite element simulations of statistical volume element (SVE) ensembles representing a PMC representative volume element (RVE). Lastly, we will present future work of a machine-learned reduced-order model between the PMC microstructure and its predicted transverse tensile stress at initial matrix failure using a Gaussian process regression (GPR) framework.

Presenting Author: Jihye Hur Georgia Institute of Technology

Presenting Author Biography: Jihye (Rachel) Hur received her B.S. in Mechanical Engineering at The University of Texas Austin and is currently a first-year Ph.D. student at the Georgia Institute of Technology. Her lab group (MINED) led by Dr. Surya Kalidindi is at the crossroads of materials science, mechanical engineering, and computational science, with research topics related to computational solid mechanics and materials informatics at varying length scales. Hur is actively engaged in research on robust quantification, design, and damage modeling of 3D woven ceramic matrix composites, and her research serves to incorporate inherently incomplete experimental information towards robust virtual mechanics studies. Hur's presentation at SSDM examines efficiently capturing complex relationships between composite microstructure and effective damage resilience, ultimately aiding an informed design of polymer matrix composites with targeted properties using machine-learned models.