Active control for drag reduction of a circular cylinder based on reinforcement learning: PIV measurements of the cylinder wake

The aim of the following paper is to experimentally investigate, using Particle Image Velocimetry, the fluid dynamic characteristics of a circular cylinder wake that has been controlled by a synthetic jet (SJ) to reduce the related aerodynamic drag. Reinforcement Learning (RL) has been applied to find the electrical signal waveshape, supplied to the loudspeaker, that allows to reduce the cylinder aerodynamic drag. Two different shapes have been tested: those obtained as Fourier series at a chosen fundamental frequency and those obtained by varying the duty cycle. The latter have been found to reduce the cylinder aerodynamic drag. Starting from this identified optimal waveshape, the use of Particle Image Velocimetry allows to obtain instantaneous two-dimensional in-plane velocity fields measurements in a plane containing the synthetic jet slot and these have been used to investigate the mean flow quantities and turbulent statistics of the phenomenon.