| Scope |
Materials design is an iterative process of identifying all feasible candidates that satisfy the design constraints and choosing the optimum which has the best target properties. The essential task is establishing the process-structure-property (P-S-P) relationships. Data-driven approaches are usually needed to explore the high-dimensional design space. Given the epistemic uncertainty inherent in simulation models and systematic errors in experiments, as well as random errors in sampling the high-dimensional space, it is challenging to construct reliable P-S-P relationships. Therefore, uncertainty quantification plays a vital role to enhance the confidence for the wide adoption of the latest data-driven materials and process design methodologies such as integrated computational materials engineering (ICME) and machine learning.
The interesting topics of this symposium include but not limited to:
- Quantifying model-form and parameter uncertainty in multiscale simulations (e.g., density functional theory, molecular dynamics, kinetic Monte Carlo, dislocation dynamics, phase field, Calphad, crystal plasticity finite-element analysis) and reduced-order models
- Quantitative methods for ICME model calibration, selection, and validation
- P-S-P surrogate modeling with statistical machine learning
- Uncertainty propagation across length and time scales
- Physics-informed machine learning to improve training efficiency and reduce prediction error
- Statistical characterization of microstructures and microstructure reconstruction
- Reliable phase equilibrium and transition state estimations with thermodynamic and first-principles methods under uncertainty
- Robust optimization with probabilistic and non-probabilistic reasoning
- Monitoring and statistical process control of manufacturing and synthesis |