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

Paper Number: 104527

104527 - Stochastic Modeling of Mechanics of Materials 

Abstract:

Multiscale computations have been widely used in the studies of mechanics of structures and materials. Computations at different scales, including quantum, atomistic, and continuum scales, can provide us with different understandings of materials and structural behavior. While smaller-scale computations provide us with more physics details, we have to turn to larger-scale computations to model materials and structures at a realistic size. A common problem when scaling up is the loss of physics. Therefore larger-scale computations cannot fully accurately describe the materials and structural behavior. Another problem arises when we compare computational results with experimental results. When we measure the mechanical behavior of a material in experiments, we will observe slightly different responses. It indicates that materials, in reality, behave in a stochastic way. However, in computations, we will get the same response of materials no matter how many times we run it. The stochastic behavior of materials can become significant for failure analysis of fracture and buckling.

In the first part, I will introduce the stochastic modeling of the uncertainties in multiscale computations. Specifically, I will talk about the method of stochastic reduced-order modeling, which is established on randomizing a reduced-order basis obtained from atomistic configuration snapshots in molecular dynamics simulations. Proper selection of hyperparameters for stochastic reduced-order modeling will allow us to model different uncertainties in multiscale computations. As a result, high-fidelity simulations can be obtained when they are compared with quantum-scale computations or experimental measurements.

In the second part, I will introduce the modeling of the stochastic mesostructures of composites and their influence on buckling behavior. The stochastic modeling was built upon a non-Gaussian random field model, which was obtained by translating a latent Gaussian field defined as the solution to a stochastic partial differential equation. The stochastic modeling allows for the construction of topology-aware spatially correlated imperfections on nonconvex domains. Our finite element modeling showed us new insights into the correlation among the manufacturing, random mesostructures, and buckling behavior of a composite shell structure.

Presenting Author: Haoran Wang Utah State University

Presenting Author Biography: Haoran Wang is an assistant professor of Mechanical & Aerospace Engineering at Utah State University (USU). Prior to joining USU, he was a postdoctoral research associate at Duke University. He obtained his Ph.D. in Aerospace Engineering from the University of Illinois at Urbana-Champaign. His research is focused on solid mechanics, multiscale computations, and data-driven modeling.

Authors: