Low-dimensional models for active control of flow separation

We study the spatiotemporal dynamics of flow separation in a planar diffuser to extract reduced-order models that may be used to guide active separation control. Proper orthogonal decomposition of numerical simulation data revealed organized dynamics of the large-scale separated structures. This result, combined with the observation (in simulations as well as laboratory experiments) of limit cycle/quasi-periodic dynamics and spatial symmetry breaking, suggests an underlying low-dimensional dynamical system. Galerkin-based and neural network-based modeling of the POD modal coefficients lead to (parameter dependent) low-dimensional models of separation dynamics. The use of such models for active separation control appears promising.