EUVIMEDEUVIMED entdecken
Crossref

Reinforcement learning-based active flow control of oscillating cylinder for drag reduction

Haokui Jiang, Shunxiang Cao

Physics of Fluids · 01.10.2023 · 35 (10)

DOI: 10.1063/5.0172081 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Physics of Fluids, online erschienen am 25.10.2023. Verfasst von Haokui Jiang und 1 weiteren Autor:innen. Die Arbeit wurde bisher 20-mal zitiert (Crossref) und nennt 67 Literaturangaben. Gefördert von: Shenzhen Peacock Plan.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This study explores the use of a reinforcement learning (RL)-based active flow control strategy to reduce the drag of a transversely oscillating cylinder confined between two walls. We incorporate wake kinematic information of the oscillating cylinder from direct numerical solution into the agent to actively adjust the oscillating amplitude. Our findings reveal two strategies that significantly improve drag reduction. First, the oscillating frequency should be chosen within the lock-in region. Second, placing probes in the active region of turbulent kinetic energy enables the full utilization of physical information for achieving more efficient and stable control. The results show that an effective control strategy can reduce the drag by 8.4% compared to the case of stationary cylinder at Re = 200. We identify three characteristics of the controlled flow that contribute to drag reduction: an elongated recirculation zone, an increased energy transport coefficient, and asymmetric cylinder oscillation. We observed that this well-behaved controlled flow can be obtained through real-time RL-based control, as it consistently attenuates the asymmetric dynamic mode decomposition modes of the flow. Furthermore, we find that the asymmetry cylinder oscillation synchronizes with the vortex shedding, indicating that the RL-based control can expand the lock-in region.

Förderung

  • Shenzhen Peacock Plan

Zitieren

APA

Jiang, H., & Cao, S. (2023). Reinforcement learning-based active flow control of oscillating cylinder for drag reduction. Physics of Fluids, 35(10). https://doi.org/10.1063/5.0172081

Vancouver

Jiang H, Cao S. Reinforcement learning-based active flow control of oscillating cylinder for drag reduction. Physics of Fluids. 2023;35(10). doi:10.1063/5.0172081

BibTeX
@article{Jiang2023,
  author = {Jiang, Haokui and Cao, Shunxiang},
  title = {Reinforcement learning-based active flow control of oscillating cylinder for drag reduction},
  journal = {Physics of Fluids},
  year = {2023},
  volume = {35},
  number = {10},
  doi = {10.1063/5.0172081}
}
Kostenlos starten →