Active flow control for drag reduction through multi-agent reinforcement learning on a turbulent cylinder at ReD=3900

This study presents novel drag reduction active-fow-control (AFC) strategies for a threedimensional cylinder immersed in a fow at a Reynolds number based on freestream velocity and cylinder diameter of ReD = 3900. The cylinder in this subcritical fow regime has been extensively studied in the litera...

ver descrição completa

Detalhes bibliográficos
Autores: Suárez Morales, Pol, Alcántara Ávila, Francisco, Miró Jané, Arnau|||0000-0002-2772-6050, Rabault, Jean, Font García, Bernat, Lehmkuhl Barba, Oriol|||0000-0002-2670-1871, Vinuesa Motilva, Ricardo
Formato: artículo
Fecha de publicación:2025
País:España
Recursos:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/426401
Acesso em linha:https://hdl.handle.net/2117/426401
https://dx.doi.org/10.1007/s10494-025-00642-x
Access Level:acceso abierto
Palavra-chave:Fluid mechanics
Drag reduction
Deep learning
Active fow control
Multiagent reinforcement learning
Àrees temàtiques de la UPC::Física::Física de fluids
Descrição
Resumo:This study presents novel drag reduction active-fow-control (AFC) strategies for a threedimensional cylinder immersed in a fow at a Reynolds number based on freestream velocity and cylinder diameter of ReD = 3900. The cylinder in this subcritical fow regime has been extensively studied in the literature and is considered a classic case of turbulent fow arising from a bluf body. The strategies presented are explored through the use of deep reinforcement learning. The cylinder is equipped with 10 independent zero-net-massfux jet pairs, distributed on the top and bottom surfaces, which defne the AFC setup. The method is based on the coupling between a computational-fuid-dynamics solver and a multi-agent reinforcement-learning (MARL) framework using the proximal-policyoptimization algorithm. This work introduces a multi-stage training approach to expand the exploration space and enhance drag reduction stabilization. By accelerating training through the exploitation of local invariants with MARL, a drag reduction of approximately 9% is achieved. The cooperative closed-loop strategy developed by the agents is sophisticated, as it utilizes a wide bandwidth of mass-fow-rate frequencies, which classical control methods are unable to match. Notably, the mass cost efciency is demonstrated to be two orders of magnitude lower than that of classical control methods reported in the literature. These developments represent a signifcant advancement in active fow control in turbulent regimes, critical for industrial applications.