|About this Abstract
||2019 TMS Annual Meeting & Exhibition
||Computational Approaches for Big Data, Artificial Intelligence and Uncertainty Quantification in Computational Materials Science
||Efficient Propagation of Uncertainty From CALPHAD to Multi-physics Phase Field Microstructure Simulations
||Pejman Honarmandi, Vahid Attari, Isaac Benson, Raymundo Arroyave, Douglas Allaire
|On-Site Speaker (Planned)
Carrying out Model Uncertainty Propagation (UP) via traditional Monte Carlo sampling requires tens of thousands of evaluations of a model, making it impractical in most computational materials science exercises. Here we present sequentially optimal sampling policy for the propagation of uncertainty from model inputs to model outputs for application to computationally expensive models, for which a budget of at most O(10) evaluations is available. Our approach is based on an optimal convergence of the input empirical distribution function to the true underlying input distribution function. The resulting approach is significantly more efficient than traditional Monte Carlo methods for uncertainty propagation. We apply this framework to the propagation of uncertainty across a phase-field model of the elasto-chemically driven microstructure evolution of a multi-phase thermoelectric system and consider uncertainty not only in the phase field parameters but also in the parameterization of the CALPHAD free energy function.
||Planned: Supplemental Proceedings volume