816 Multi-Fidelity Simulation and Modeling in Computational Science and Engineering

Xueyu Zhu, University of Iowa
Akil Narayan, University of Utah
John Jakeman, Sandia National Lab
Multi-fidelity methods can improve the efficiency and robustness of predictions by leveraging competing models with varying degrees of trust. The abstract concept of "fidelity" -- faithfulness to physics, experimental observations, and/or idealized mathematical formulations -- arises because different simulation suites utilize different discretization types and scales and make dissimilar simplifications of underlying physics. This mini-symposium aims to highlight recent advances and developments in algorithms that make optimal use of models with differing fidelities.
Algorithms of interest include those that make efficient use of multiscale hierarchies, leverage inexpensive lower fidelity information to boost accuracy of expensive high-fidelity information, manage and analyze computational resource allocation across scales and models, provide certifiable error guarantees for multi-fidelity model fusion, assimilate experimental and/or observational data into simulation predictions, or provide novel algorithmic strategies for multi-fidelity model management.
This minisymposium will highlight modern challenges such as computational budget allocations among models, identification and learning of model hierarchy, and efficient synthesis of model predictions. We expect the presentations in this minisymposium to catalyze research thrusts that aim to address these issues, and to provide a forum that encourages interaction of interdisciplinary researchers.