Bayesian optimization for active flow control

A key question in flow control is that of the design optimal controllers when space high-dimensional and experimental or computational budget limited. We address this formidable challenge using a particular flavor machine learning present first application Bayesian optimization to open-loop for fluid flows. consider range acquisition functions, including recently introduced output-informed criteria Blanchard Sapsis (2021), evaluate performance algorithm two iconic configurations active control: computationally, with drag reduction fluidic pinball; experimentally, mixing enhancement turbulent jet. For these flows, we find identifies at fraction cost other strategies considered previous studies. also provides, as by-product optimization, surrogate model latent function, which can be leveraged paint complete picture landscape. The proposed methodology used virtually any complex and, therefore, has significant implications an industrial scale.