Machine learning strategies applied to the control of a fluidic pinball

Abstract

The wake stabilization of a triangular cluster of three rotating cylinders is investigated. Experiments are performed at Reynolds number Re ∼2200. Flow control is realized using rotating cylinders spanning the wind-tunnel height. The cylinders are individually connected to identical brushless DC motors. Two-component planar particle image velocimetry measurements and constant temperature hot-wire anemometry were used to characterize the flow without and with actuation. Main open-loop configurations are studied and different controlled flow topologies are identified. Machine learning control is then implemented for the optimization of the flow control performance. Linear genetic algorithms are used here as the optimization technique for the open-loop constant speed-actuators. Two different cost functions J are considered targeting either drag reduction or wake symmetrization. The functions are estimated based on the velocity from three hot-wire sensors in the wake. It is shown that the machine learning approach is an effective strategy for controlling the wake characteristics. More significantly, the results show that machine learning strategies can reveal unanticipated solutions or parameter relations, in addition to being a tool for optimizing searches in large parameter spaces.