About this Abstract |
Meeting |
2020 TMS Annual Meeting & Exhibition
|
Symposium
|
Materials Design Approaches and Experiences V
|
Presentation Title |
Alloy Design Through Sequential Learning |
Author(s) |
James Edward Saal |
On-Site Speaker (Planned) |
James Edward Saal |
Abstract Scope |
The state-of-the-art of alloy design relies on physical models and databases to map out the processing-structure-property relationships responsible for materials performance. Successful application of this approach to alloys has been demonstrated in steels, superalloys, and other structural alloys with an extensive history of development. However, alloy classes without such history lack accurate models or populated datasets to enable design. In such applications, materials informatics can accelerate the design workflow. Machine learning models can be trained on existing, sparsely populated data to derive correlation between processing-structure-properties in the absence of mechanistic understanding. An iterative process called sequential learning uses these machine learning models to identify the best experiments to perform to reduce model uncertainty and/or identify the highest-performing materials. Citrine’s experience using sequential learning to design alloys will be discussed, including the use cases of high entropy alloys (HEAs) for structural applications and thermoelectric half-Heusler solid solutions. |
Proceedings Inclusion? |
Planned: Supplemental Proceedings volume |