Machine Learning Applications in Materials Data and Imaging
Department of Materials Science and Engineering, University of Wisconsin – Madison, WI USA
Machine learning methods are impacting almost every aspect of our lives, and science and engineering are no exception. Here we briefly discuss some of the areas where machine learning is changing how we do Materials Science and Engineering, and some specific applications we have pursued in our group. The first application is to modeling diffusion coefficients and demonstrates the ability of elemental properties to capture key aspects of the physics of atomic transport and predict diffusion coefficients. The second application is to modeling irradiation hardening effects in steels for nuclear applications and demonstrates that simply knowing composition and irradiation conditions can allow quantitative prediction of irradiation response. Both these applications make use of kernel regression approaches and I will discuss some of the challenges of these and related approaches in assessing model errors and domain of applicability. The third application is to automating the reading of electron microscopy images to assess damage from irradiation in nuclear steels and demonstrates that machine vision tools can successfully count defects and measure their dimensions. This work suggests that many image analysis activities in materials characterization may soon be largely automated.
Brief Biographical Sketch for Dane Morgan
Dane Morgan obtained a Ph.D. in Physics from U.C. Berkeley in 1998, was a Postdoctoral Researcher and Research Scientist at MIT until 2004, and is now a Professor in Materials Science and Engineering and co-director of the Wisconsin Materials Institute at the University of Wisconsin, Madison. His work combines thermostatistics, thermokinetics, and informatics analysis with atomic scale calculations to understand and predict materials properties. A major focus of Morgan’s work is energy applications, including fuel cells, batteries, and nuclear materials, but he also works in the areas of high-pressure geoscience and defect properties in semiconductors. Morgan has done extensive consulting work for industry and in 2011 served as vice president of research at Pellion Technologies, a startup energy technology company. Morgan is presently training or has graduated/trained over 60 graduate students and postdoctoral researchers, and worked with approximately 100 undergraduates in research. He is the Harvey D. Spangler Professor of Engineering and a University of Wisconsin Vilas Scholar, received the University of Wisconsin Award for Mentoring Undergraduates in Research, Scholarly and Creative Activities and the Harvey Spangler Award for Innovative Teaching and Learning Practices for his undergraduate mentoring activities, has won multiple top paper awards, and has published over 250 papers in materials science.