GEORGIA INSTITUTE OF TECHNOLOGY

Under the provisions of the regulations for the degree

DOCTOR OF PHILOSOPHY

on Monday, April 25, 2022

11:00 AM

via

BlueJeans Video Conferencing
https://bluejeans.com/787872949/5450

will be held the

DISSERTATION PROPOSAL DEFENSE

for

Sven Voigt

"A Knowledge Graph for Materials Informatics"

Committee Members:

Prof. Surya Kalidindi, Advisor, CSE/ME/MSE

Prof. David McDowell, ME/MSE

Prof. Josh Kacher, MSE

Prof. Sham Navathe, CoC

Zach Trautt, Ph.D., National Institute of Standards and Technology (NIST)

Abstract:

A knowledge graph (KG) helps meet the findable, accessible, interoperable, and reusable (FAIR) principles of data by capturing the important conceptual information needed to understand data such that it can be reused. In particular, this work proposes the materials informatics KG for capturing process-structure-property (PSP) relationships between data that is currently stored in a fragmented manner in order to make it FAIR. The KG is composed of three new components: an ontological model designed to digitally capture PSP materials knowledge, digital object architecture for KGs, and specific artificial intelligence (AI) tools that infer additional PSP knowledge. In addition to existing materials ontological models, the new ontological model will make PSP relationships explicit. In addition to current KG technology, which represents textual information, the new KG incorporates emerging digital object architecture concepts to capture knowledge in data files, which may also possess many heterogeneous file formats. In addition to existing AI tools, which are built for general KGs or are built for fragmented data, new AI tools purposely built to take advantage of the combination of the graphical structure and the specific materials PSP data are used to aggregate existing knowledge and extract new knowledge. The proposal is on the development and the validation of this KG, where validation includes demonstrating the ability to capture sufficient PSP knowledge, aggregating PSP knowledge using AI tools, and inferring the likelihood of PSP characteristics for a new material specimen given existing knowledge. These new technologies enable capturing and inferring much more knowledge than other materials data solutions and are combined with powerful KG query (e.g., search) capabilities to make materials data much more FAIR.