DIFFER seminar: Data-driven modeling in interconnected dynamical systems

Abstract: Dynamical systems that are currently addressed in engineering and science and for which monitoring, control, decision-making and diagnostics challenges are present, typically become more and more complex, large-scale and interconnected. Many of the operational challenges require accurate models that need to be build, often based on physical first principles. Operational and experimental data play a paramount role in either validating the models, adapting them to changing circumstances, or even building the dynamic models from scratch. This requires techniques for handling data in interconnected dynamic systems, referred to as dynamic networks, and for turning this data into accurate model information on the networks. Over the past fifteen years a research domain in data-driven aspects in dynamic networks has been developed, and in this seminar a brief overview of this will be given. Topics that will be addressed include: modeling concepts of either directed and undirected networks, related to the two different modeling frameworks as present inSIMULINK andSIMSCAPE; network identifiability; allocation of sensors and actuators for identifying either full networks or subnetworks, and related identification methods.