You're invited to attend the talk

"Predicting Microstructure-Sensitive Mechanical Behavior of Materials by Integrating Physics-Based Modeling with Machine Learning"
 

by
 

Prof. Ashley Spear

Assistant Professor | Mechanical Engineering
The University of Utah

Thursday, October 17
3 - 4 p.m.
Guggenheim 442

About the Talk: 
The emergence of a data-driven science paradigm in the mechanics of materials community has been simultaneously motivated and enabled by recent, unprecedented access to multidimensional and multimodal data derived from sophisticated experiments, computational simulations, and combinations thereof. This talk will focus on specific examples in which we have exploited data derived from high-fidelity, physics-based models and/or from 3D experimental characterization to train machine-learning algorithms to predict complex material phenomena. In one example, we hypothesize that micromechanical fields in a cyclically loaded, uncracked polycrystal encode the eventual 3D path of a crack, which we test by training machine-learning models using data derived from crystal-plasticity simulations combined with experimental observations collected at the Advanced Photon Source. In a second example, we test the performance of different types of machine-learning models (Ridge regression, random forests, and convolutional neural networks) in terms of their ability to predict microstructure-sensitive mechanical properties throughout additively manufactured (AM) metal build domains. The models are trained using data generated from high-fidelity, multi-physics numerical simulations of the AM process, resulting microstructure, and effective mechanical properties. In each example, the primary aim of using machine learning is to massively expedite predictions of complex physical phenomena. In this context, machine learning can open up new opportunities to search high-dimensional materials-design spaces, make rapid predictions of microstructure-sensitive failure, and even discover fundamental relationships between raw image data and mechanical behavior for cases in which the governing equations are unknown or intractable to solve.

About the Speaker: 
Dr. Ashley Spear directs the Multiscale Mechanics & Materials Laboratory at the University of Utah, where her group specializes in integrating experiments, physics-based modeling, and data-driven science to examine three-dimensional deformation, fatigue, and fracture in structural materials. Spear received her B.S. in Architectural Engineering from the University of Wyoming and her Ph.D. in Civil Engineering from Cornell University. She has received numerous teaching awards, the Young Investigator Award from the Air Force Office of Scientific Research, the TMS Early Career Faculty Fellow Award, and the CAREER Award from the National Science Foundation.