The impacts of many important technologies are limited by the availability of better-performing materials. One factor limiting the ability of engineers to develop better materials is the speed at which they can search through possible formulations and processing schemes. Recently, machine learning algorithms have emerged as a possible route to reusing existing materials data to guide the design process. In this thesis, we discuss work towards addressing three major challenges in the use of machine learning in materials engineering. First, we implemented an automated toolkit for solving crystal structures and use it to improve the quality of an existing materials database. Second, we developed general-purpose methods for creating machine learning models from materials data, which will simplify and accelerate the development of new models. Third, we created open-source software for making these machine learning techniques more readily-accessible to the materials community. Along with addressing these challenges, we also demonstrate how machine learning can be applied to optimize existing and discover new Bulk Metallic Glass alloys. It is our vision that the methods developed in this work will help enable the application of machine learning to a wider variety of problems and, potentially, be used to improve materials employed in many different technologies.

Last modified

• 03/13/2018

Creator

• Logan Timothy Ward

DOI

• https://doi.org/10.21985/N27T2Q

Subject

• Materials Science and Engineering

Keyword

• Materials Science
• Crystal Structure Solution
• Metallic Glasses
• Density Functional Theory
• Machine Learning

Date created

• 2017-01-01

Resource type

• Dissertation

Rights statement

• In Copyright