Materials informatics of structural metals and alloys
Abstract: Engineering performance of materials for advanced load-bearing applications are strongly dependent on their internal structure (microstructure). Quantitative microstructure--property relationships are key for efficient design of new high-performance materials. Machine learning is a promising approach for accelerated data-driven development of materials. Complex and hierarchical nature of the microstructure as well as scarce data often hinder the use of off-the-shelf machine learning methods. In this talk, I will discuss new approaches to quantification of microstructures for enabling machine learning of microstructure-property relationships, including graph representation of grain structures of alloys.
