Closed-loop flow separation control using the deep q network over airfoil

Abstract

The effectiveness of closed-loop flow control using a deep Q network (DQN), such as deep reinforcement learning algorithm, was experimentally researched for a NACA0015 airfoil. The system used a dielectric barrier discharge plasma actuator as the flow control device, and the experiment was conducted at the chord Reynolds number of 6.3 × 104 . The closed-loop control system selected a nondimensional burst frequency of the actuator by analyzing the time series of the surface pressure data. The neural network of the DQN was sequentially trained at the angles of attack of 12 and 15 deg. As a result, the closed-loop control successfully generated a higher control gain compared with the open-loop control using a fixed burst frequency. Particularly, at 15 deg there was a significant control gain that could not be obtained by the open-loop control. The closed-loop control keeps the flow attached and preserves it for a longer time by periodically switching the actuator on and off.