Deep reinforcement learning control strategy for synthetic jets on airfoils

Aiming at insufficient lift and poor aerodynamic stability in low Reynolds number (Re) operation conditions of airfoil represented by low-speed aircraft in near space, this study adopted deep reinforcement learning (DRL) method to implement active flow control on the synthetic jet of National Advisory Committee for Aeronautics (NACA) 0012 airfoil at Re<1500. The problem of sharp fluctuation of airfoil lift resistance under low Reynolds number is solved, and the mechanism of deep reinforcement learning control strategy is explored. On this basis, the deep reinforcement learning control strategy of the synthetic jet at different positions is further analyzed. The numerical simulation results show that the deep reinforcement learning control strategy can achieve more stable control than the traditional synthetic jet control method. The resultant jet partially inhibits the flow separation and reduces the lift fluctuation. The shear layer is deflected to the airfoil, the pressure difference between the upper and lower surfaces of the airfoil is increased, and the lift force is increased. It is also found that when the actuator is located on the lower surface of the trailing edge, the control strategy tends to increase the jet velocity. When the actuator is located on the upper surface, the control strategy shows a better ability to improve the lift–drag ratio.