Shocks Under Control: Taming Transonic Compressible Flow over an RAE2822 Airfoil with Deep Reinforcement Learning

TL;DR

This work tackles transonic SBLI on a 2D RAE2822 airfoil at $R_e=50{,}000$ by coupling a high-fidelity compressible Navier–Stokes solver (5th-order spectral DG, SSPRK54, AMR) with a deep reinforcement learning agent that controls a three-by-three array of synthetic jets. Using a 678-sensor pressure state and a PPO-based policy (with a TD3 off-policy comparison in extended tests), the study demonstrates that DRL can learn coordinated actuation to suppress shock-induced separation and reduce oscillations, achieving up to a $\sim 25\%$ drag reduction and lift enhancements up to $\sim 196\%$, with lift-to-drag improvements exceeding $200\%$ in some cases. Actuation frequency and reward design significantly influence performance, with intermediate frequencies and lift-preserving rewards yielding robust gains, while very high frequencies can degrade controllability. The results establish DRL-driven AFC on high-fidelity solvers as a viable path toward practical drag reduction and efficiency improvements in transonic configurations, and point to future work on sensor placement, regime variation, and 3D extension for real-world applicability.

Abstract

Active flow control of compressible transonic shock-boundary layer interactions over a two-dimensional RAE2822 airfoil at Re = 50,000 is investigated using deep reinforcement learning (DRL). The flow field exhibits highly unsteady dynamics, including complex shock-boundary layer interactions, shock oscillations, and the generation of Kutta waves from the trailing edge. A high-fidelity CFD solver, employing a fifth-order spectral discontinuous Galerkin scheme in space and a strong-stability-preserving Runge-Kutta (5,4) method in time, together with adaptive mesh refinement capability, is used to obtain the accurate flow field. Synthetic jet actuation is employed to manipulate these unsteady flow features, while the DRL agent autonomously discovers effective control strategies through direct interaction with high-fidelity compressible flow simulations. The trained controllers effectively mitigate shock-induced separation, suppress unsteady oscillations, and manipulate aerodynamic forces under transonic conditions. In the first set of experiments, aimed at both drag reduction and lift enhancement, the DRL-based control reduces the average drag coefficient by 13.78% and increases lift by 131.18%, thereby improving the lift-to-drag ratio by 121.52%, which underscores its potential for managing complex flow dynamics. In the second set, targeting drag reduction while maintaining lift, the DRL-based control achieves a 25.62% reduction in drag and a substantial 196.30% increase in lift, accompanied by markedly diminished oscillations. In this case, the lift-to-drag ratio improves by 220.26%.

Figures (18)

• Figure 1: Compressible transonic flow over the RAE2822 2D airfoil with synthetic jet actuation. Three synthetic jets are depicted on both the upper and lower surfaces of the airfoil.
• Figure 2: Density contours illustrating flow features around the airfoil. (a) Density contours showing Kutta waves interacting with shock waves and the wake region. (b) Density contours computed with adaptive mesh refinement, highlighting enhanced resolution in critical regions. (c) Zoomed view near the airfoil surface, revealing boundary layer separation.
• Figure 3: Drag and Lift coefficients for 5th, 6th, and 7th order polynomials.
• Figure 4: The kinetic energy spectra for polynomial orders $p = 5, 6,$ and $7$ at $t = 0$, $1$, and $2$ seconds, illustrating the convergence of the resolved scales.
• Figure 5: Compressible transonic flow over the RAE2822 2D airfoil with synthetic jet actuation. Three synthetic jets are depicted on both the upper and lower surfaces of the airfoil.