Recently, Deep Reinforcement Learning (DRL) has rapidly advanced Active Flow Control (AFC) for blunt bodies, while presenting new possibilities in AFC of airfoils. The present work aims to effectively control flow around an airfoil using two 0.5% chord-length jets, with their maximum velocity capped at four times the freestream velocity. A Soft Actor-Critic (SAC)-based framework is used to control jets’ velocity to enhance aerodynamic performance at Re = 1000. Compared with previous studies, the agent employs completely distinct strategies under varying angles of attack α . At α = 10°, the agent employs near-constant action with upstream suction and downstream blowing; at α = 20°, it utilizes a periodic action with a dominant frequency approximately four times that of the coefficient fluctuations, breaking down the leading-edge vortex; at α = 30°, the agent adopts complex action significantly reducing the coefficient fluctuations’ frequency. The results demonstrate the effectiveness of the SAC-based framework, which achieves a drag reduction by 15.2% at α = 20°. Meanwhile, the performance across different α presents the adaptability of the algorithm, indicating the potential of DRL for developing more intelligent and adaptive strategies that can effectively manage complex flow phenomena.
Zhang et al. (Wed,) studied this question.