About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Computational Materials for Qualification and Certification
|
| Presentation Title |
Robust and Efficient Design of Additively Manufactured Alloys by Integrating Uncertainty Quantification and Modeling Using Generative AI |
| Author(s) |
Hasan Al Jame, Bo Ni, Mohadeseh Taheri-Mousavi |
| On-Site Speaker (Planned) |
Hasan Al Jame |
| Abstract Scope |
Additive manufacturing often suffers from processing-induced compositional variability, resulting in uncertainties that can compromise the mechanical reliability of critical components. To address this challenge, we present an uncertainty-informed inverse design framework to identify alloy composition ranges that deliver reliable properties within a pre-defined, allowable uncertainty limit. Taking yield strength as a demonstration, we coupled CALPHAD simulations with physics-based modeling to predict its distribution across the standardized compositional space and performed global sensitivity analysis to reveal the critical elements driving the uncertainty. Leveraging this information, Bayesian optimization was applied to determine their optimal concentration ranges to achieve targeted strength within the desired uncertainty threshold. We automate this workflow of forward prediction, uncertainty quantification, and inverse design using a multi-agent generative AI technique. Our framework is expected to provide valuable guidelines for robust yet efficient AM alloy designs with controlled uncertainty and expedite the qualification and certification process for mission-critical components. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Additive Manufacturing, ICME, Modeling and Simulation |