Vortex shedding suppression in elliptical cylinder via reinforcement learning

Abstract

Flow control of bluff bodies plays a critical role in engineering applications. In this study, deep reinforcement learning (DRL) is employed to develop flow control strategies for the flow past an elliptical cylinder confined between two walls. The primary objective is to investigate the feasibility of achieving multi-objective flow control for an elliptical cylinder with varying aspect ratios ($Ar$), while maintaining low control energy input. DRL training results demonstrate that for an elliptical cylinder with larger $Ar$, the control strategy effectively reduces drag, minimizes lift fluctuations, and completely suppresses vortex shedding, all while maintaining low external energy consumption. Conversely, decreasing the $Ar$ compromises the effectiveness of multi-objective control, even when greater energy input is applied. Through detailed physical analysis, the coupling effect between the blockage ratio ($\beta$) and $Ar$ is identified as a limiting factor for vortex shedding suppression and wake stabilization. At lower values of $\beta$, the control strategy successfully achieves multi-objective optimization for elliptical cylinders across the entire range of $Ar$. Although balancing energy efficiency and control performance remains challenging for highly slender cylinders, the proposed DRL strategy still achieves effective vortex shedding suppression. This work highlights the potential of DRL-based control strategies to effectively stabilize wake flows around slender bluff bodies, with an explicit emphasis on maintaining energy efficiency.