Data-driven modeling of fluids

-
Clarence Rowley, Princeton University
Fine Hall 214

In-Person Talk (Princeton ID Holders Only)

Fluid flows can be extraordinarily complex, and even turbulent, yet often there is structure lying within the apparent complexity. Understanding this structure can help explain observed physical phenomena, and can help with the design of control strategies in situations where one would like to change the natural state of a flow. This talk addresses techniques for obtaining simple, approximate models for fluid flows, using data from simulations or experiments. We discuss a number of methods, including principal component analysis, balanced truncation, and Koopman operator methods, and focus on a new method for optimizing projections of nonlinear systems. We apply these techniques to several flows with complex behavior, including a transitional channel flow and an axisymmetric jet.