| Topic: | Fast Prediction of Thermo-Mechanical Properties of Materials using Machine Learning Potentials |
|---|---|
| Speaker: | Dr Satish Kumar Professor, G. W. Woodruff School of Mechanical Engineering Georgia Institute of Technology |
| Date: | Friday, 6 December 2024 |
| Time: | 2.00pm to 3.00pm |
| Venue: | Seminar Room EA-06-02 Block EA, Level 6, College of Design & Engineering, NUS (Click here for location map.) |
| Host: | A/Prof Koh Yee Kan |
| Empirical interatomic potentials are much faster than first-principles methods and can scale to much larger atomic systems, but are less accurate, difficult to develop, and lack transferability. Machine learning (ML) approaches offer new opportunities for the development of low-computational cost surrogate models trained to the expensive Density Functional Theory (DFT) computations. The value of a ML potential lies in how efficiently and accurately it learns the complex physics underlying the atomic interactions. We developed a voxelized atomic structure representation of the physical volume around the atom of interest, which is equivalent to capturing the atomic neighborhood as a digital 3D image. This is ideal as an input to convolutional neural networks (CNNs) for implicit feature engineering. The calibrated convolutional layers of the CNN serve as a complex, nonlinear mapping from the voxelized representation of the atomic neighborhood to a low-dimensional feature space that is finally used to predict the net atomic force. We develop voxelized atomic structure potential (VASt), which is capable of predicting highly accurate interatomic forces of different electronic materials. We use VASt potential to compute thermal conductivity with the Boltzmann Transport Equation in a highly accurate and computationally efficient manner. VASt potential is trained using data generated from ab initio molecular dynamics simulations. Our calculations suggest VASt potential is capable of accurately reproducing the thermal conductivity of Si over a range of isotropic strain. Next, we integrate VASt potential in a homogenization framework for fast prediction of thermo-mechanical properties of high entropy alloys as a function of chemical composition. |
 Dr. Satish Kumar is currently Professor in George W. Woodruff School of Mechanical Engineering at Georgia Tech. Prior to joining Georgia Tech in 2009 as an Assistant Professor, he worked at IBM Corporation where he was responsible for the thermal management of electronic devices. Kumar received his Ph.D. in Mechanical Engineering and M.S. degree in Electrical and Computer Engineering from Purdue University, West Lafayette in 2007; and B.Tech. degree in Mechanical Engineering from the Indian Institute of Technology, Guwahati in 2001. His research interests are in electro-thermal transport study in electronic devices and materials, e.g., heterogeneous electronics, electric motors, etc. He is author or co-author of over 160 journal or conference publications. Dr. Kumar is an ASME Fellow and recipient of 2005 Purdue Research Foundation Fellowship, 2012 Summer Faculty Fellow from Air Force Research Lab, 2014 Sigma Xi Young Faculty Award, 2014 DARPA Young Faculty Award, 2017 Woodruff Faculty Fellow, and 2020 ASME K-16 Clock Award. Dr. Kumar has been named as Frank H. Neely Professor, at Georgia Tech. |
(All are welcome to attend.)
Event Details
Date & Time
6 December 2024, Friday
14:00 PM - 15:00 PM
14:00 PM - 15:00 PM
Venue
Seminar Room EA-06-02, Block EA, Level 6, College of Design & Engineering, NUS
Organiser
Department of Mechanical Engineering