Speaker: Richard LeSar, Lynn Gleason Professor of Interdisciplinary Engineering, Department of Materials Science and Engineering, Iowa State University
Title: Materials Analytics: Getting the Most out of Complex Datasets
Abstract:
Materials research is all about generating data, whether from experiment or modeling and simulation. This data is only useful to the extent we can extract information about the system under study. For some types of data, it is easy to extract that information. For example, the relation between two variables (e.g., stress versus strain) can easily be represented by a simple plot. More general two-variable properties (e.g., strength vs elastic modulus) can be captured for a wide variety of material types using an Ashby plot. However, data is often much more complex, often involving many data types. Extracting information from such multidimensional datasets requires techniques that are not as commonly used in materials research as they should be. In this talk will will focus on a number of methods from the field of data analytics, providing a tutorial on techniques for dimension reduction in multidimensional datasets (e.g., principal component analysis), for correlation between variables in multidimensional datasets (e.g., canonical correlation analysis), for representation of complex microstructures for relating structures to properties (e.g., n-point correlation functions), and others. The goal of the talk is to provide sufficient background, as well as links to commonly available tools, to show how these methods can be helpful in extracting the needed information out of the complex datasets that make up materials research.