Data-Driven Balancing Transformation for Predictive Model Order Reduction

Explore data-driven balancing transformation techniques for predictive model order reduction in this hour-long webinar. Delve into the system theoretic foundations of balanced truncation and its data-driven variations, focusing on the eigensystem realization algorithm (ERA). Examine the predictive performance of ERA reduced-order models (ROMs) in complex scenarios, including a benchmark reactive flow problem and aero-acoustic response prediction. Learn about tangential interpolation methods for reducing offline costs in multi-input multi-output systems. Discover strategies to address computational bottlenecks in systems with numerous input-output channels, such as output domain decomposition and low-fidelity gappy POD approaches. Gain insights from speaker Elnaz Rezaian, a Postdoctoral Fellow at the University of Michigan, on surrogate modeling and non-intrusive model reduction in complex systems using system identification and machine learning techniques.

DDPS | 'Data-driven balancing transformation for predictive model order reduction'