About this Abstract |
| Meeting |
2027 TMS Annual Meeting & Exhibition
|
| Symposium
|
Advanced Welding, Joining, and Bonding of Structural Steels
|
| Presentation Title |
Physics-Informed Inverse Design of Welding Filler Materials for Low-Alloy Steels with High Strength, Ductility and Toughness |
| Author(s) |
Gyunghun Yu, Mohadeseh Taheri-Mousavi |
| On-Site Speaker (Planned) |
Gyunghun Yu |
| Abstract Scope |
Welding filler alloy design for low-alloy steels requires balancing composition, processing parameters, microstructure, and mechanical properties within a high-dimensional design space. We present a data-driven framework for forward prediction and inverse design of low-alloy steel weld metals using exploratory data analytics, machine learning, and Bayesian optimization (BO). Composition and process variables were analyzed together with weld metal properties, including strength, ductility, and toughness, to extract physically meaningful trends. To better capture hardenability effects, a carbon-equivalent feature was introduced and found to correlate strongly with mechanical-property trends. A surrogate model was developed to predict weld properties from composition and process variables and coupled with BO to search candidate compositions for target property combinations. The inverse design results were further examined against experimental trends to assess physical consistency, establishing a practical framework for accelerated design of low-alloy steel welding filler materials. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Joining, ICME, Machine Learning |