About this Abstract |
| Meeting |
2027 TMS Annual Meeting & Exhibition
|
| Symposium
|
Verification, Calibration, and Validation Approaches for Mechanical Modeling of Metals
|
| Presentation Title |
Principles, Mechanisms, Predictors: What It Takes to Trust a Structure–Property Model |
| Author(s) |
Raymundo Arroyave |
| On-Site Speaker (Planned) |
Raymundo Arroyave |
| Abstract Scope |
Understanding the physical world happens at several levels. One holds a principle—alloys with smaller grains tend to be stronger. A theory then commits to a mechanism (grain-boundary–dislocation interactions) and condenses it into a phenomenological model. Beyond sit models agnostic to mechanism but tuned to predict. None of these are first-principles models — each must be calibrated against data before being confronted with the world. Three questions follow: (1) is a single model consistent with the data, (2) is one model superior to others, and (3) what metrics should adjudicate? I will tackle these using a recently acquired dataset connecting chemistry, processing, microstructure, and mechanical properties in an FCC HEA space. A second question sits beneath: the model that predicts best may rank worst — a distinction that matters when models move from helping us understand the world to acting in it, as in closed-loop alloy design. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, Machine Learning, ICME |