|About this Abstract
||1st World Congress on High Entropy Alloys (HEA 2019)
||High Entropy Alloys 2019
||Machine Learning-aided Accelerated Discovery of HEA for Turbomachinery Applications
||Soumalya Sarkar, Kenneth D Smith, John A Sharon, Ryan M Deacon, GV Srinivasan
|On-Site Speaker (Planned)
The feasible space of possible HEA, demonstrating novel thermo-mechanical properties necessary for high-efficiency turbomachinery applications, is still largely unexplored. Although some of the state-of-the-art high-throughput computational approaches have been able to scan through a large equi-molar HEA space or various HEA systems with a narrow range of composition variation under a limited set of property requirements, they haven’t yet scaled up to a larger HEA space with high resolution of composition variation, which is necessary to discover non-obvious HEAs with multiple competing properties. This paper presents a machine learning-based framework which is able to converge on a set of HEA candidates given a large set of design objectives and constraints. The proposed approach demonstrates scalability to comprehensive HEA space exploration even while receiving data from variably expensive physics-based thermo-mechanical models. Experimental data from test coupons can also be incorporated in this framework to aid to the reliability of overall framework.