Uncertainty Quantification in Data-Centric Simulations Workshop

This one-day intensive workshop provides a broad survey of mathematical methods for uncertainty quantification (UQ). Participants will learn about the key sources of uncertainty in modeling complex dynamic systems (model form, parameters, noise measurement, etc.) and how to address them using state-of-the-art techniques. By the end of the day, attendees will be familiar with and have practical exposure to:

• Fundamental probability and statistics concepts for quantifying uncertainty
• Spatial statistics techniques (e.g. Gaussian process modeling/Kriging) for building data-driven surrogate models
• Monte Carlo simulation methods for uncertainty propagation and risk analysis
• Global and local sensitivity analysis to determine influential parameters
• Surrogate modeling approaches to reduce computational cost (including polynomial chaos expansions and other alternatives to brute-force Monte Carlo)
• Data assimilation methods (Kalman filtering) for updating computational models with observational or real-time data

Workshop Instructor

Daniel Tartakovsky, Professor of Energy Science Engineering