|About this Abstract
||2020 TMS Annual Meeting & Exhibition
||High Entropy Alloys VIII
||A Machine Learning Model for Alloy Design
||Zhaohan Zhang, Mu Li, Katharine Flores, Rohan Mishra
|On-Site Speaker (Planned)
Developing fast and accurate methods to promote alloy discovery is of practical interest, especially with the vast composition space offered by multi-principal element alloys (MPEAs). While density-functional-theory (DFT)-based methods have accelerated design of binary and ternary alloys, they are not amenable for rapidly screening the vast combinatorial space of MPEAs. We develop a machine-learning model for predicting the DFT-calculated formation enthalpy of alloys and use it to identify stable alloys. The model uses easily accessible elemental properties as descriptors and has a mean absolute error (MAE) of ∼ 6 meV/atom for binary alloys. We use the model to successfully identify new binary intermetallics that are subsequently confirmed using DFT and experiments. Model trained with binary intermetallics can predict formation enthalpy of ternary intermetallics with an MAE as good as DFT calculation. We further apply it to MPEAs to predict the formation of single-phase solid solutions with bcc and fcc structures.
||Planned: Supplemental Proceedings volume; Planned: Supplemental Proceedings volume