| Affiliations: | Aggie Research Mentoring Program |
| Project Leader: | Brent Vela
brentvela@tamu.edu Materials Science & Engineering |
| Faculty Mentor: | Raymundo Arroyave, Ph.D. |
|
Meeting Times:
|
Friday noon (discuss and update meeting) and Tuesday noon (silent work, together/ office hours) |
|
Team Size:
|
5
|
| Open Spots: | 0 |
|
Special Opportunities:
|
co-authorship, learn how to work on supercomputing clusters, learn data-science (transferable skill)
|
|
Team Needs:
|
Knowledge of the following Python packages; Pandas, Numpy, Matplotlib, Seaborn Basic machine learning knowledge is a plus. Knowledge of Linux is a plus. Basic materials science (metals) is a plus. |
|
Description:
|
High entropy alloys (HEAs) are a novel class of alloys that have garnered attention in metallurgical communities for their superior properties. HEAs are mixtures of elements at near-equi-atomic compositions as opposed to conventional alloys where a single base element is alloyed with minor additions of other elements. There are many ways to combine 4 or more elements at different portions in an alloy. In fact, the number of potential HEA candidates is combinatorically vast. We seek to explore the HEA space in search of alloys for various applications, however this requires experimental validation. These experiments are often very expensive. Therefore, in order to avoid trail-and-error approaches, we seek to use a data-driven approach to design HEAs for performance (e.g. strength, phase-stability, oxidation resistance), processability (e.g. 3D printability), and sustainability (e.g. low cost, avoid scarce elements). The project will have 3 main aspects: 1) Bayesian design-of-experiment (DOE) methods will be developed that will help researchers intelligently pick the next best alloy to scale up for experiment. 2) Advanced machine learning models will be developed that will fuse experimental knowledge and prior physics-based knowledge into a single model. This will create physics-informed model that should be less data-hungry than conventional machine learning models. 3) New design “indicators” will be implemented from the existing literature. These indicators are simple models that are used to screen candidate HEAs, filtering out alloys that are likely not desirable, and leaving behind alloys that likely meet certain specifications. For more information about what this project will consist of, see the papers here: https://scholar.google.com/citations?hl=en&user=MRdi5GkAAAA |