In-person MSE Colloquium April 1: Prasanna Balachandran, Data-Driven Materials Design and Its Synergy with Mechanistic Models in Materials Informatics

University of Virginia

Abstract

Data-Driven Materials Design and Its Synergy with Mechanistic Models in Materials Informatics

In the rapidly growing field of materials informatics, there is a constant interest in discovering novel and creative ways to incorporate domain knowledge and uncover previously unseen patterns existing in the data. In our research group, we are currently exploring two distinct routes to leverage artificial intelligence-based approaches in materials design. The first approach takes a data-driven motivation. The domain knowledge is incorporated in the form of domain-inspired descriptors. Machine learning methods are integrated with uncertainty quantification schemes, optimal learning methods and an “oracle” to efficiently navigate the vast search space in an adaptive or iterative manner. One of the expected outcomes from such an adaptive learning framework is a predictive black-box surrogate model that is perceived to capture the complexity of the structure-property relationships with sufficient accuracy. More recently, our group has been integrating novel post hoc model interpretability methods to peek inside the trained/validated black-box surrogate models and explain the predictions for each observation or a related collection of observations in the training data. The second approach combines experimental data with numerical or analytical models within the framework of Bayesian inference. Here, we rely on mechanistic models that are meticulously formulated through observations to validate specific hypotheses. We are exploring simple strategies that harness the promise of machine learning algorithms to overcome the limitations of mechanistic models in predictive studies. In this talk, I will focus on examples that highlight the potential of both approaches in accelerating new materials design.
 

Bio

Prasanna Balachandran is currently an Assistant Professor in the Department of Materials Science and Engineering with a joint appointment in the Department of Mechanical and Aerospace Engineering at the University of Virginia (UVA). He earned his Bachelors’ Degree in Metallurgical Engineering from Anna University, India in 2007 and Ph.D. in Materials Science and Engineering from Iowa State University, USA in 2011. Prior to joining UVA in December 2017, he spent three and half years as a postdoctoral research associate in the Theoretical Division at Los Alamos National Laboratory (LANL), USA and two years as a postdoctoral research associate at Drexel University, USA. His research interests are interdisciplinary spanning diverse areas such as crystal symmetry, first-principles based density functional theory calculations, and information science methods for accelerating the design and discovery of new materials. Prasanna thoroughly enjoys introducing the concepts of artificial intelligence and numerical methods to materials science undergraduate students.

This presentation is in-person with a Zoom option for virtual attendance.

Zoom

https://osu.zoom.us/s/99442121610

Password: 351334