kth.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Deep Reinforcement Learning for Flow Control Exploits Different Physics for Increasing Reynolds Number Regimes
CMT—Motores Térmicos, Universitat Politècnica de València, 46022 Valencia, Spain.ORCID iD: 0000-0002-7909-4569
Barcelona Super Computing Center—Centro Nacional de Supercomputación (BSC-CNS), 08034 Barcelona, Spain;FLOW, Engineering Mechanics, KTH Royal Institute of Technology, 114 28 Stockholm, Sweden.ORCID iD: 0000-0002-6031-5536
KTH, School of Engineering Sciences (SCI), Engineering Mechanics. KTH, School of Engineering Sciences (SCI), Centres, Linné Flow Center, FLOW.ORCID iD: 0000-0003-0704-6100
Barcelona Super Computing Center—Centro Nacional de Supercomputación (BSC-CNS), 08034 Barcelona, Spain.ORCID iD: 0000-0002-2772-6050
Show others and affiliations
2022 (English)In: Actuators, E-ISSN 2076-0825, Vol. 11, no 12, article id 359Article in journal (Refereed) Published
Abstract [en]

The increase in emissions associated with aviation requires deeper research into novel sensing and flow-control strategies to obtain improved aerodynamic performances. In this context, data-driven methods are suitable for exploring new approaches to control the flow and develop more efficient strategies. Deep artificial neural networks (ANNs) used together with reinforcement learning, i.e., deep reinforcement learning (DRL), are receiving more attention due to their capabilities of controlling complex problems in multiple areas. In particular, these techniques have been recently used to solve problems related to flow control. In this work, an ANN trained through a DRL agent, coupled with the numerical solver Alya, is used to perform active flow control. The Tensorforce library was used to apply DRL to the simulated flow. Two-dimensional simulations of the flow around a cylinder were conducted and an active control based on two jets located on the walls of the cylinder was considered. By gathering information from the flow surrounding the cylinder, the ANN agent is able to learn through proximal policy optimization (PPO) effective control strategies for the jets, leading to a significant drag reduction. Furthermore, the agent needs to account for the coupled effects of the friction- and pressure-drag components, as well as the interaction between the two boundary layers on both sides of the cylinder and the wake. In the present work, a Reynolds number range beyond those previously considered was studied and compared with results obtained using classical flow-control methods. Significantly different forms of nature in the control strategies were identified by the DRL as the Reynolds number Re increased. On the one hand, for Re & LE;1000, the classical control strategy based on an opposition control relative to the wake oscillation was obtained. On the other hand, for Re=2000, the new strategy consisted of energization of the boundary layers and the separation area, which modulated the flow separation and reduced the drag in a fashion similar to that of the drag crisis, through a high-frequency actuation. A cross-application of agents was performed for a flow at Re=2000, obtaining similar results in terms of the drag reduction with the agents trained at Re=1000 and 2000. The fact that two different strategies yielded the same performance made us question whether this Reynolds number regime (Re=2000) belongs to a transition towards a nature-different flow, which would only admits a high-frequency actuation strategy to obtain the drag reduction. At the same time, this finding allows for the application of ANNs trained at lower Reynolds numbers, but are comparable in nature, saving computational resources.

Place, publisher, year, edition, pages
MDPI AG , 2022. Vol. 11, no 12, article id 359
Keywords [en]
numerical simulation, wake dynamics, flow control, machine learning, deep reinforcement learning
National Category
Fluid Mechanics and Acoustics
Identifiers
URN: urn:nbn:se:kth:diva-356269DOI: 10.3390/act11120359ISI: 000900414000001Scopus ID: 2-s2.0-85144726353OAI: oai:DiVA.org:kth-356269DiVA, id: diva2:1912782
Note

QC 20241217

Available from: 2024-11-13 Created: 2024-11-13 Last updated: 2024-12-17Bibliographically approved
In thesis
1.
The record could not be found. The reason may be that the record is no longer available or you may have typed in a wrong id in the address field.
2. Multi-agent reinforcement learning for enhanced turbulence control in bluff bodies
Open this publication in new window or tab >>Multi-agent reinforcement learning for enhanced turbulence control in bluff bodies
2024 (English)Licentiate thesis, comprehensive summary (Other academic)
Alternative title[sv]
Multi-agent förstärkningsinlärning för förbättrad turbulensreglering i bluffkroppar
Abstract [en]

This licentiate thesis explores the application of deep reinforcement learning (DRL) to flow control in bluff bodies, focusing on reducing drag forces in infinite cylinders. The research spans a range of flow conditions, from laminar to fully turbulent, aiming to advance the state-of-the-art in DRL by exploring novel scenarios not yet covered in the fluid-mechanics literature. Our focus is on the flow around cylinders in two and three dimensions, over a range of Reynolds numbers Re_D based on freestream velocity U and cylinder diameter D. We first consider a single-agent reinforcement learning (SARL) approach using the proximal-policy optimization (PPO) algorithm, coupled with the Alya numerical solver. This approach led to significant drag reductions of 20% and 17.7% for Re_D = 1000 and 2000, respectively, in a two-dimensional (2D) setting. The framework was designed for deployment on high-performance computers, enabling large-scale training with synchronized numerical simulations.

Next, we focused on three-dimensional (3D) cylinders, where spanwise instabilities emerge for Re_D > 250. Drawing inspiration from studies such as Williamson (1996) and findings from Tang et al. (2020), we explored strategies for Re_D = 100 to 400 with a multi-agent reinforcement learning (MARL) framework. This approach focused on local invariants, using multiple jets across the top and bottom surfaces. The MARL framework successfully reduced drag by 21% and 16.5% for Re_D = 300 and 400, respectively, outperforming periodic-control strategies by 10 percentage points and doubling efficiency.

Finally, the framework was tested in a fully turbulent environment at Re_D = 3900, a well-established case in the literature. Despite the significant computational challenges and complex flow structures, the MARL approach delivered significant results, with an 8.3% drag reduction and reducing the mass flow used in the actuation by two orders of magnitude compared with Kim & Choi (2005). Across these studies, the drag-reduction mechanisms learned by the agents involve altering the wake topology to attenuate and move the location of the Reynolds-stresses maximum values upstream, focusing on enlarging the recirculation bubble and reducing pressure drag.

Abstract [sv]

Denna licentiatavhandling utforskar möjligheten att använda förstärkningsin-lärning (DRL) för strömningskontroll kring trubbiga kroppar där speciellt fokus ligger på att minska motståndskrafterna på oändliga cylindrar. En rad strömningsförhållanden undersöks från laminärt till fullt utvecklad turbulent strömning. Målet är att bygga vidare på den senaste utvecklingen inom DRL genom att utforska nya strömningsförhållanden som ännu inte har behandlats inom strömningsmekaniken. Vårt fokus ligger på strömning runt cylindrar i två och tre dimensioner samt över ett spektrum av Reynolds-tal Re_D baserat på fri strömningshastigheten U och cylinderns diameter D. I den första delen av avhandlingen utvecklades en enskild agent-förstärkningsinlärningsmetod med proximal policy optimization kopplad till den numeriska lösaren Alya. Denna metod ledde till betydande minskningar i motståndskrafterna på 20% och 17,7% för Re_D = 1000 och 2000 i en tvådimensionell (2D) miljö. Detta ramverk för kontroll, numerisk simulering och analys utformades för att köras på högpresterande datorer vilket möjliggjorde storskalig träning av nätverket med synkroniserade numeriska simuleringar. Därefter fokuserade vi på tredimensionella (3D) cylindrar, där instabiliteter längs cylinderaxeln uppträder vid Re_D > 250. Inspirerade av studier som Williamson (1996) och resultat från Tang et al. (2020), undersökte vi strategier för Re_D = 100 till 400 med ett multiagent-förstärkningsinlärningsramverk (MARL). Denna metod fokuserade på lokala invariansprinciper och använde flera jetstrålar över cylinderns övre och nedre ytor för kontroll av strömningen. MARL-ramverket minskade motståndet med 21% respektive 16,5% för Re_D = 300 och 400 och överträffade periodiska kontrollstrategier med 10 procentenheter och fördubblade effektiviteten. Slutligen testades ramverket på turbulent strömning vid Re_D = 3900, ett välkänt fall i litteraturen. Trots beräkningsutmaningar och komplexa flödesstrukturer minskade MARL motståndet med 8,3% och massflödet med två storleksordningar jämfört med Kim & Choi (2005). Gemensamt för våra studier är att de motståndsminskande mekanismer som lärts av agenterna involverar att förändra strömningsvakens topologi för att dämpa och flytta Reynolds-stressernas maximala värden uppströms. Detta leder till förstora återcirkulationsbubblan och minska tryckmotståndet.

Place, publisher, year, edition, pages
Stockholm: KTH Royal Institute of Technology, 2024. p. 124
Series
TRITA-SCI-FOU ; 2024:53
Keywords
Machine learning, active flow control, deep reinforcement learning, fluid mechanics, turbulence, Fluidmekanik, laminär-turbulent övergång, aktiv flödeskontroll, dragreduktion, maskininlärning, djup förstärkningsinlärning
National Category
Fluid Mechanics and Acoustics Engineering and Technology
Research subject
Engineering Mechanics
Identifiers
urn:nbn:se:kth:diva-356281 (URN)978-91-8106-103-1 (ISBN)
Presentation
2024-12-05, F2, Kungliga Tekniska Högskolan, Lindstedtsvägen 26 & 28, Stockholm, 13:00 (English)
Opponent
Supervisors
Funder
EU, European Research Council, grant no.2021-CoG-101043998, DEEPCONTROL
Note

QC 241114

Available from: 2024-11-14 Created: 2024-11-13 Last updated: 2024-12-19Bibliographically approved

Open Access in DiVA

fulltext(11226 kB)33 downloads
File information
File name FULLTEXT01.pdfFile size 11226 kBChecksum SHA-512
5e43493e6c7e2dc8bcbd94881a8e0d6f3045bed748025bc2a85f54c42b7ab2c22f96cebae51ee193e2bed32d930f9e03b2f7089df6776d7f4a0eaa8377f0cac4
Type fulltextMimetype application/pdf

Other links

Publisher's full textScopus

Authority records

Alcantara-Avila, FranciscoVinuesa, Ricardo

Search in DiVA

By author/editor
Varela, PauSuárez, PolAlcantara-Avila, FranciscoMiró, ArnauRabault, JeanFont, BernatGarcía-Cuevas, Luis MiguelLehmkuhl, OriolVinuesa, Ricardo
By organisation
Engineering MechanicsLinné Flow Center, FLOW
Fluid Mechanics and Acoustics

Search outside of DiVA

GoogleGoogle Scholar
Total: 33 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

doi
urn-nbn

Altmetric score

doi
urn-nbn
Total: 233 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf