About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Algorithms Development in Materials Science and Engineering
|
| Presentation Title |
A Data-Driven Probabilistic Framework for Quantifying Uncertainty in Predicting the Yield Strength of Precipitate-Strengthened FCC Alloys |
| Author(s) |
Jing Luo, Ahamed Ali N, Jaafar El-Awady |
| On-Site Speaker (Planned) |
Jing Luo |
| Abstract Scope |
Advances in automated alloy design require predictive frameworks that not only estimate mechanical properties but also rigorously quantify uncertainty. To address this, we introduce a parameter-free probabilistic framework for predicting the yield strength of precipitate-strengthened FCC alloys. The approach combines: (1) a physics-based model that generates probability distributions of stress from precipitate strengthening based on classical theory and calibrated by the inherent variability in experimental data; (2) a probabilistic model connecting dislocation density to stress; and (3) a data-driven model trained on experimental data to map microstructures features to dislocation density. By integrating these components, the framework can predict the yield strength distribution of polycrystalline without empirical fitting. It also isolates and quantifies uncertainty contributions from microstructural variability, providing actionable insights for uncertainty reduction. This work demonstrates the power of uniting physical theory, data science, and statistical inference to enhance prediction accuracy and accelerate the design of next-generation structural alloys. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, Modeling and Simulation, Machine Learning |