Dynamic Network Analysis for Robust Uncertainty Management

Andrzej Banaszuk, United Technologies Research Center

We will present an approach to acceleration of numerical simulations, model reduction, and quantification of uncertainty in large multi-scale nonlinear networks of interconnected dynamic components. In this approach a large network is decomposed into subcomponents using spectral graph theory. Operator theory and geometric dynamics methods are used to analyze quantification of uncertainty in subcomponents. We will also show how this approach could enable model-based analysis and design of robust aerospace and building systems by managing interconnections between components. This research is conducted by DARPA Robust Uncertainty Management team including UTRC, UCSB, Caltech, Stanford, Yale, Georgia Tech, and Princeton.