Detection of flow regime transitions using dynamic mode decomposition and moving horizon estimation

The spatial and time behaviors of fluid flows at different Reynolds numbers and free-stream turbulence intensity levels are studied by combining dynamic mode decomposition (DMD) and moving horizon estimation to detect flow-regime transitions. In more detail, the norm of residuals provided by DMD when processing successive snapshots of the flow velocity field shows a trend that is identified by means of a moving horizon estimator based on a switching model. This allows detecting the change from stable to unstable flow regimes, which in turn enables to extract modes, frequencies, and growth rates of com- plex structures such as vortices, characterizing the fluid flow in the spatial and temporal domains. Different cases of experimental measurements given by a particle image velocimetry are analyzed to recognize the complexity of the underlying flow physics, while showing the effectiveness of the proposed approach.