Stabilizing the square cylinder wake using deep reinforcement learning for different jet locations

Abstract

We mitigate the drag and lift forces of a square cylinder at three Reynolds numbers of 500, 1000, and 2000 using deep reinforcement learning (DRL) and two different positions of a zero flux jet actuator couple based on computational fluid dynamics simulations. The jet actuators are symmetrically deployed at the leading and trailing corners of the square cylinder and tested at those three Reynolds numbers. Surface pressure probes monitor the flow state featuring a practical engineering solution as compared to velocity sensors in the wake. For each jet actuator position and Reynolds number, the feedback law is optimized using the powerful capabilities of DRL with the pressure signals as control input and the jet velocities as control output. Under leading jet control, the mean drag coefficient is reduced by 44.4%, 60.6%, and 57.8% for increasing Reynolds numbers. In addition, the lift fluctuation is reduced by 85.9%, 82%, and 86.2%, demonstrating a substantial stabilization of the wake. In contrast, the optimized trailing jet control performs much worse. This study highlights the significance of the location of zero-net-mass-flux actuation and may guide the practical application of DRL-based active flow control of square cylinders.