When
Student: Sheila Whitman, Program in Applied Mathematics
Title: New Microstructure Representations for Process–Structure–Property Modeling of Alloys
Advisors: Marat Latypov, Materials Science & Engineering Dept.
Location: ENR2, Room S395 | Zoom link: https://arizona.zoom.us/j/83509060198, Meeting ID: 835 0906 0198
Abstract: New materials are necessary to meet global energy, transportation, defense, and healthcare needs. Accelerated materials design relies on understanding how the composition and processing conditions determine the structure of the material and how the structure dictates the properties of a material. Machine learning models can capture these relationships, but their success relies on quantitative representation of the microstructure. In this dissertation, new quantitative microstructure representations for processing--structure--property (PSP) relationships of structural alloys are presented. First, I introduce a new quantitative description, spatially-resolved chord length distributions, for characterizing complex, non-uniform microstructures. Next, I demonstrate that foundational vision transformers can learn effective microstructure representations directly from images of the microstructure without segmentation or materials-specific training. Finally, I present a generative microstructure model developed for recycled aluminum alloys. Evaluation of these generative models rely on new representation approaches and complete the PSP chains for computational design of sustainable alloys.