Open Access
Subscription or Fee Access
On the Stochastic Characterization and Prediction of Data-Driven Multi-Scale Materials Models for Composites
Abstract
The work presented is this paper is carried out towards a project on Integrated computational materials engineering for the development of carbon fiber composite for light vehicles. Here, we focus on highlighting selected mathematical and computational aspects pertaining to the stochastic modeling of composites across a hierarchy of length scales. These aspects have been addressed to control the construction of validated structural performance predictors for composites as functions of microscopic parametric descriptors. We show the significance of developing appropriate prior models of fine-scale properties for the characterization process, i.e. their impact on how to devise reliable macroscopic stochastic predictors and how to infer updated microscopic stochastic models. Specifically, we delineate on the process of modeling the moduli of fiber in tension and compression as well as modeling their statistical dependency in view of macroscopic data-driven model updating process. Tensile and flexural physical experiments at the coupon level are employed for this purpose. The characterization of key quantities of interests by means of polynomial chaos expansion has been requisite for developing explicit stochastic coupling across the hierarchy of length-scales as part of the characterization and the verification processes.
DOI
10.12783/asc35/34976
10.12783/asc35/34976