Deep Reinforcement Learning Control of Cylinder Flow Using Rotary Oscillations at Low Reynolds Number

Standard

Deep Reinforcement Learning Control of Cylinder Flow Using Rotary Oscillations at Low Reynolds Number. / Tokarev, Mikhail; Palkin, Egor; Mullyadzhanov, Rustam.

в: Energies, Том 13, № 22, 5920, 02.11.2020.

Результаты исследований: Научные публикации в периодических изданиях › статья › Рецензирование

BibTeX

@article{9321714ec9184294a4ab8960f21e4a90,

title = "Deep Reinforcement Learning Control of Cylinder Flow Using Rotary Oscillations at Low Reynolds Number",

abstract = "We apply deep reinforcement learning to active closed-loop control of a two-dimensional flow over a cylinder oscillating around its axis with a time-dependent angular velocity representing the only control parameter. Experimenting with the angular velocity, the neural network is able to devise a control strategy based on low frequency harmonic oscillations with some additional modulations to stabilize the K{\'a}rm{\'a}n vortex street at a low Reynolds number Re = 100. We examine the convergence issue for two reward functions showing that later epoch number does not always guarantee a better result. The performance of the controller provide the drag reduction of 14% or 16% depending on the employed reward function. The additional efforts are very low as the maximum amplitude of the angular velocity is equal to 8% of the incoming flow in the first case while the latter reward function returns an impressive 0.8% rotation amplitude which is comparable with the state-of-the-art adjoint optimization results. A detailed comparison with a flow controlled by harmonic oscillations with fixed amplitude and frequency is presented, highlighting the benefits of a feedback loop.",

keywords = "flow control, ANN, DRL, CIRCULAR-CYLINDER, GENETIC ALGORITHM, DYNAMICS, WAKE, Flow control",

author = "Mikhail Tokarev and Egor Palkin and Rustam Mullyadzhanov",

year = "2020",

month = nov,

day = "2",

doi = "10.3390/en13225920",

language = "English",

volume = "13",

journal = "Energies",

issn = "1996-1073",

publisher = "MDPI AG",

number = "22",

}

RIS

TY - JOUR

T1 - Deep Reinforcement Learning Control of Cylinder Flow Using Rotary Oscillations at Low Reynolds Number

AU - Tokarev, Mikhail

AU - Palkin, Egor

AU - Mullyadzhanov, Rustam

PY - 2020/11/2

Y1 - 2020/11/2

N2 - We apply deep reinforcement learning to active closed-loop control of a two-dimensional flow over a cylinder oscillating around its axis with a time-dependent angular velocity representing the only control parameter. Experimenting with the angular velocity, the neural network is able to devise a control strategy based on low frequency harmonic oscillations with some additional modulations to stabilize the Kármán vortex street at a low Reynolds number Re = 100. We examine the convergence issue for two reward functions showing that later epoch number does not always guarantee a better result. The performance of the controller provide the drag reduction of 14% or 16% depending on the employed reward function. The additional efforts are very low as the maximum amplitude of the angular velocity is equal to 8% of the incoming flow in the first case while the latter reward function returns an impressive 0.8% rotation amplitude which is comparable with the state-of-the-art adjoint optimization results. A detailed comparison with a flow controlled by harmonic oscillations with fixed amplitude and frequency is presented, highlighting the benefits of a feedback loop.

AB - We apply deep reinforcement learning to active closed-loop control of a two-dimensional flow over a cylinder oscillating around its axis with a time-dependent angular velocity representing the only control parameter. Experimenting with the angular velocity, the neural network is able to devise a control strategy based on low frequency harmonic oscillations with some additional modulations to stabilize the Kármán vortex street at a low Reynolds number Re = 100. We examine the convergence issue for two reward functions showing that later epoch number does not always guarantee a better result. The performance of the controller provide the drag reduction of 14% or 16% depending on the employed reward function. The additional efforts are very low as the maximum amplitude of the angular velocity is equal to 8% of the incoming flow in the first case while the latter reward function returns an impressive 0.8% rotation amplitude which is comparable with the state-of-the-art adjoint optimization results. A detailed comparison with a flow controlled by harmonic oscillations with fixed amplitude and frequency is presented, highlighting the benefits of a feedback loop.

KW - flow control

KW - ANN

KW - DRL

KW - CIRCULAR-CYLINDER

KW - GENETIC ALGORITHM

KW - DYNAMICS

KW - WAKE

KW - Flow control

UR - http://www.scopus.com/inward/record.url?scp=85104291980&partnerID=8YFLogxK

U2 - 10.3390/en13225920

DO - 10.3390/en13225920

M3 - Article

VL - 13

JO - Energies

JF - Energies

SN - 1996-1073

IS - 22

M1 - 5920

ER -

ID: 27913868