|About this Abstract
||2016 TMS Annual Meeting & Exhibition
||Computational Methods for Uncertainty Quantification, Model Validation, and Stochastic Predictions
||Functional Uncertainty Quantification for Multi-fidelity and Multi-scale Simulations
||Sam Reeve, Alejandro Strachan
|On-Site Speaker (Planned)
Uncertainty quantification (UQ) continues to increase in importance in materials simulation, as does the predictive power and the ability to make decisions using those simulations. We describe functional uncertainty quantification methods which determine uncertainty in a given quantity of interest (QoI) arising from uncertainty in the input constitutive functions. While most UQ focuses on error from uncertain input parameters, this methodology focuses on error from the functional forms themselves. Numerical calculation of the functional derivative with respect to a low-fidelity (coarse-scale) input function makes possible error correction of a QoI for a multi-fidelity (or multi-scale) simulation. Alternatively, high-fidelity (fine-scale) additional simulations can be ranked based on the product of the functional sensitivity and discrepancy to maximally improve the overall accuracy. Application has included multi-fidelity molecular dynamics and multi-scale crystal plasticity with finite element simulations.
||Planned: A print-only volume