About this Abstract |
Meeting |
2025 TMS Annual Meeting & Exhibition
|
Symposium
|
Algorithms Development in Materials Science and Engineering
|
Presentation Title |
Bayesian Classification for Constraining the Design of Compositionally Graded Alloys (CGAs) |
Author(s) |
James Hanagan, Eli Norris, Maryam Ghotbi, Brent Vela, Raymundo Arróyave |
On-Site Speaker (Planned) |
James Hanagan |
Abstract Scope |
CGAs allow for optimizing alloy properties only in regions of the part where they are needed, as opposed to optimizing a single alloy—with all the associated property trade-offs—to withstand every condition the part will be exposed to. The challenge with CGA design is avoiding deleterious compositions when grading between two selected endpoints. Constraining compositions with model-predicted properties prior to designing the CGA is helpful for avoiding bad compositions, but what happens when models fail to predict behavior accurately? In this work, a classifier is proposed that can constrain the alloy design space using model input as a prior belief which can then be updated with experimental data. This enables an iterative approach to CGA development where previous designs can better inform subsequent ones. The classifier, which relies on Gaussian process regression with informative priors, will be discussed along with its application to CGA design. |
Proceedings Inclusion? |
Planned: |
Keywords |
Additive Manufacturing, ICME, Machine Learning |