Deep learning fluid flow reconstruction around arbitrary two-dimensional objects from sparse sensors using conformal mappings

The usage of neural networks (NNs) for flow reconstruction (FR) tasks from a limited number of sensors is attracting strong research interest owing to NNs’ ability to replicate high-dimensional relationships. Trained on a single flow case for a given Reynolds number or over a reduced range of Reynol...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:AIP advances 2022-04, Vol.12 (4), p.045126-045126-22
Hauptverfasser: Özbay, Ali Girayhan, Laizet, Sylvain
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page 045126-22
container_issue 4
container_start_page 045126
container_title AIP advances
container_volume 12
creator Özbay, Ali Girayhan
Laizet, Sylvain
description The usage of neural networks (NNs) for flow reconstruction (FR) tasks from a limited number of sensors is attracting strong research interest owing to NNs’ ability to replicate high-dimensional relationships. Trained on a single flow case for a given Reynolds number or over a reduced range of Reynolds numbers, these models are unfortunately not able to handle flows around different objects without re-training. We propose a new framework called Spatial Multi-Geometry FR (SMGFR) task, capable of reconstructing fluid flows around different two-dimensional objects without re-training, mapping the computational domain as an annulus. Different NNs for different sensor setups (where information about the flow is collected) are trained with high-fidelity simulation data for a Reynolds number equal to ∼300 for 64 objects randomly generated using Bezier curves. The performance of the models and sensor setups is then assessed for the flow around 16 unseen objects. It is shown that our mapping approach improves percentage errors by up to 15% in SMGFR when compared to a more conventional approach where the models are trained on a Cartesian grid and achieves errors under 3%, 10%, and 30% for predictions of pressure, velocity, and vorticity fields, respectively. Finally, SMGFR is extended to predictions of snapshots in the future, introducing the Spatiotemporal MGFR (STMGFR) task. A novel approach is developed for STMGFR involving splitting deep neural networks into a spatial and a temporal component. We demonstrate that this approach is able to reproduce, in time and in space, the main features of flows around arbitrary objects.
doi_str_mv 10.1063/5.0087488
format Article
fullrecord <record><control><sourceid>proquest_scita</sourceid><recordid>TN_cdi_scitation_primary_10_1063_5_0087488</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><doaj_id>oai_doaj_org_article_d9c85f3166724668915780ee4bc769bd</doaj_id><sourcerecordid>2655283453</sourcerecordid><originalsourceid>FETCH-LOGICAL-c428t-abdbda65cda9adc7040e3e9b85760bd25aeb13280024884daa360acb4bf9858a3</originalsourceid><addsrcrecordid>eNqdkctq3TAQQE1JoSG5i_6BoKsWnMh6WV6WNGkDgW7StRg9HHSxLXckN3SRf49ub2izrhYaMXM4Ymaa5n1HLzqq-KW8oFT3Qus3zSnrpG45Y-rk1ftds8t5T-sRQ0e1OG2evoSwkikALnF5IOO0RV_v9EgwuLTkgpsrMS0EMG2Lr8HGgoC_SXlMrY9zWHItw0SS3QdXMhkxzSSvgDmQXKsJM9nyQV59Y8K5sjOsa83k8-btCFMOu5d41vy4ub6_-tbeff96e_X5rnWC6dKC9daDks7DAN71VNDAw2C17BW1nkkItuNMU8pq88IDcEXBWWHHQUsN_Ky5PXp9gr1ZMc61A5Mgmj-JhA8GsEQ3BeMHp-XIO6V6JpTSQyd7TUMQ1vVqsL66PhxdK6afW8jF7NOGdQLZMCUl01xIXqmPR8phyhnD-PfXjprDsow0L8uq7Kcjm10scJj2_8G_Ev4DzepH_gxcUaVT</addsrcrecordid><sourcetype>Open Website</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2655283453</pqid></control><display><type>article</type><title>Deep learning fluid flow reconstruction around arbitrary two-dimensional objects from sparse sensors using conformal mappings</title><source>DOAJ Directory of Open Access Journals</source><source>Elektronische Zeitschriftenbibliothek - Frei zugängliche E-Journals</source><source>Alma/SFX Local Collection</source><source>Free Full-Text Journals in Chemistry</source><creator>Özbay, Ali Girayhan ; Laizet, Sylvain</creator><creatorcontrib>Özbay, Ali Girayhan ; Laizet, Sylvain</creatorcontrib><description>The usage of neural networks (NNs) for flow reconstruction (FR) tasks from a limited number of sensors is attracting strong research interest owing to NNs’ ability to replicate high-dimensional relationships. Trained on a single flow case for a given Reynolds number or over a reduced range of Reynolds numbers, these models are unfortunately not able to handle flows around different objects without re-training. We propose a new framework called Spatial Multi-Geometry FR (SMGFR) task, capable of reconstructing fluid flows around different two-dimensional objects without re-training, mapping the computational domain as an annulus. Different NNs for different sensor setups (where information about the flow is collected) are trained with high-fidelity simulation data for a Reynolds number equal to ∼300 for 64 objects randomly generated using Bezier curves. The performance of the models and sensor setups is then assessed for the flow around 16 unseen objects. It is shown that our mapping approach improves percentage errors by up to 15% in SMGFR when compared to a more conventional approach where the models are trained on a Cartesian grid and achieves errors under 3%, 10%, and 30% for predictions of pressure, velocity, and vorticity fields, respectively. Finally, SMGFR is extended to predictions of snapshots in the future, introducing the Spatiotemporal MGFR (STMGFR) task. A novel approach is developed for STMGFR involving splitting deep neural networks into a spatial and a temporal component. We demonstrate that this approach is able to reproduce, in time and in space, the main features of flows around arbitrary objects.</description><identifier>ISSN: 2158-3226</identifier><identifier>EISSN: 2158-3226</identifier><identifier>DOI: 10.1063/5.0087488</identifier><identifier>CODEN: AAIDBI</identifier><language>eng</language><publisher>Melville: American Institute of Physics</publisher><subject>Artificial neural networks ; Cartesian coordinates ; Conformal mapping ; Curves ; Errors ; Fluid dynamics ; Fluid flow ; Machine learning ; Neural networks ; Object recognition ; Reconstruction ; Reynolds number ; Sensors ; Training ; Two dimensional flow ; Vorticity</subject><ispartof>AIP advances, 2022-04, Vol.12 (4), p.045126-045126-22</ispartof><rights>Author(s)</rights><rights>2022 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).</rights><lds50>peer_reviewed</lds50><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c428t-abdbda65cda9adc7040e3e9b85760bd25aeb13280024884daa360acb4bf9858a3</citedby><cites>FETCH-LOGICAL-c428t-abdbda65cda9adc7040e3e9b85760bd25aeb13280024884daa360acb4bf9858a3</cites><orcidid>0000-0003-0346-0662 ; 0000-0002-2100-847X</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>314,780,784,864,2102,27924,27925</link.rule.ids></links><search><creatorcontrib>Özbay, Ali Girayhan</creatorcontrib><creatorcontrib>Laizet, Sylvain</creatorcontrib><title>Deep learning fluid flow reconstruction around arbitrary two-dimensional objects from sparse sensors using conformal mappings</title><title>AIP advances</title><description>The usage of neural networks (NNs) for flow reconstruction (FR) tasks from a limited number of sensors is attracting strong research interest owing to NNs’ ability to replicate high-dimensional relationships. Trained on a single flow case for a given Reynolds number or over a reduced range of Reynolds numbers, these models are unfortunately not able to handle flows around different objects without re-training. We propose a new framework called Spatial Multi-Geometry FR (SMGFR) task, capable of reconstructing fluid flows around different two-dimensional objects without re-training, mapping the computational domain as an annulus. Different NNs for different sensor setups (where information about the flow is collected) are trained with high-fidelity simulation data for a Reynolds number equal to ∼300 for 64 objects randomly generated using Bezier curves. The performance of the models and sensor setups is then assessed for the flow around 16 unseen objects. It is shown that our mapping approach improves percentage errors by up to 15% in SMGFR when compared to a more conventional approach where the models are trained on a Cartesian grid and achieves errors under 3%, 10%, and 30% for predictions of pressure, velocity, and vorticity fields, respectively. Finally, SMGFR is extended to predictions of snapshots in the future, introducing the Spatiotemporal MGFR (STMGFR) task. A novel approach is developed for STMGFR involving splitting deep neural networks into a spatial and a temporal component. We demonstrate that this approach is able to reproduce, in time and in space, the main features of flows around arbitrary objects.</description><subject>Artificial neural networks</subject><subject>Cartesian coordinates</subject><subject>Conformal mapping</subject><subject>Curves</subject><subject>Errors</subject><subject>Fluid dynamics</subject><subject>Fluid flow</subject><subject>Machine learning</subject><subject>Neural networks</subject><subject>Object recognition</subject><subject>Reconstruction</subject><subject>Reynolds number</subject><subject>Sensors</subject><subject>Training</subject><subject>Two dimensional flow</subject><subject>Vorticity</subject><issn>2158-3226</issn><issn>2158-3226</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2022</creationdate><recordtype>article</recordtype><sourceid>DOA</sourceid><recordid>eNqdkctq3TAQQE1JoSG5i_6BoKsWnMh6WV6WNGkDgW7StRg9HHSxLXckN3SRf49ub2izrhYaMXM4Ymaa5n1HLzqq-KW8oFT3Qus3zSnrpG45Y-rk1ftds8t5T-sRQ0e1OG2evoSwkikALnF5IOO0RV_v9EgwuLTkgpsrMS0EMG2Lr8HGgoC_SXlMrY9zWHItw0SS3QdXMhkxzSSvgDmQXKsJM9nyQV59Y8K5sjOsa83k8-btCFMOu5d41vy4ub6_-tbeff96e_X5rnWC6dKC9daDks7DAN71VNDAw2C17BW1nkkItuNMU8pq88IDcEXBWWHHQUsN_Ky5PXp9gr1ZMc61A5Mgmj-JhA8GsEQ3BeMHp-XIO6V6JpTSQyd7TUMQ1vVqsL66PhxdK6afW8jF7NOGdQLZMCUl01xIXqmPR8phyhnD-PfXjprDsow0L8uq7Kcjm10scJj2_8G_Ev4DzepH_gxcUaVT</recordid><startdate>20220401</startdate><enddate>20220401</enddate><creator>Özbay, Ali Girayhan</creator><creator>Laizet, Sylvain</creator><general>American Institute of Physics</general><general>AIP Publishing LLC</general><scope>AJDQP</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>8FD</scope><scope>H8D</scope><scope>L7M</scope><scope>DOA</scope><orcidid>https://orcid.org/0000-0003-0346-0662</orcidid><orcidid>https://orcid.org/0000-0002-2100-847X</orcidid></search><sort><creationdate>20220401</creationdate><title>Deep learning fluid flow reconstruction around arbitrary two-dimensional objects from sparse sensors using conformal mappings</title><author>Özbay, Ali Girayhan ; Laizet, Sylvain</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c428t-abdbda65cda9adc7040e3e9b85760bd25aeb13280024884daa360acb4bf9858a3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2022</creationdate><topic>Artificial neural networks</topic><topic>Cartesian coordinates</topic><topic>Conformal mapping</topic><topic>Curves</topic><topic>Errors</topic><topic>Fluid dynamics</topic><topic>Fluid flow</topic><topic>Machine learning</topic><topic>Neural networks</topic><topic>Object recognition</topic><topic>Reconstruction</topic><topic>Reynolds number</topic><topic>Sensors</topic><topic>Training</topic><topic>Two dimensional flow</topic><topic>Vorticity</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Özbay, Ali Girayhan</creatorcontrib><creatorcontrib>Laizet, Sylvain</creatorcontrib><collection>AIP Open Access Journals</collection><collection>CrossRef</collection><collection>Technology Research Database</collection><collection>Aerospace Database</collection><collection>Advanced Technologies Database with Aerospace</collection><collection>DOAJ Directory of Open Access Journals</collection><jtitle>AIP advances</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Özbay, Ali Girayhan</au><au>Laizet, Sylvain</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Deep learning fluid flow reconstruction around arbitrary two-dimensional objects from sparse sensors using conformal mappings</atitle><jtitle>AIP advances</jtitle><date>2022-04-01</date><risdate>2022</risdate><volume>12</volume><issue>4</issue><spage>045126</spage><epage>045126-22</epage><pages>045126-045126-22</pages><issn>2158-3226</issn><eissn>2158-3226</eissn><coden>AAIDBI</coden><abstract>The usage of neural networks (NNs) for flow reconstruction (FR) tasks from a limited number of sensors is attracting strong research interest owing to NNs’ ability to replicate high-dimensional relationships. Trained on a single flow case for a given Reynolds number or over a reduced range of Reynolds numbers, these models are unfortunately not able to handle flows around different objects without re-training. We propose a new framework called Spatial Multi-Geometry FR (SMGFR) task, capable of reconstructing fluid flows around different two-dimensional objects without re-training, mapping the computational domain as an annulus. Different NNs for different sensor setups (where information about the flow is collected) are trained with high-fidelity simulation data for a Reynolds number equal to ∼300 for 64 objects randomly generated using Bezier curves. The performance of the models and sensor setups is then assessed for the flow around 16 unseen objects. It is shown that our mapping approach improves percentage errors by up to 15% in SMGFR when compared to a more conventional approach where the models are trained on a Cartesian grid and achieves errors under 3%, 10%, and 30% for predictions of pressure, velocity, and vorticity fields, respectively. Finally, SMGFR is extended to predictions of snapshots in the future, introducing the Spatiotemporal MGFR (STMGFR) task. A novel approach is developed for STMGFR involving splitting deep neural networks into a spatial and a temporal component. We demonstrate that this approach is able to reproduce, in time and in space, the main features of flows around arbitrary objects.</abstract><cop>Melville</cop><pub>American Institute of Physics</pub><doi>10.1063/5.0087488</doi><tpages>22</tpages><orcidid>https://orcid.org/0000-0003-0346-0662</orcidid><orcidid>https://orcid.org/0000-0002-2100-847X</orcidid><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier ISSN: 2158-3226
ispartof AIP advances, 2022-04, Vol.12 (4), p.045126-045126-22
issn 2158-3226
2158-3226
language eng
recordid cdi_scitation_primary_10_1063_5_0087488
source DOAJ Directory of Open Access Journals; Elektronische Zeitschriftenbibliothek - Frei zugängliche E-Journals; Alma/SFX Local Collection; Free Full-Text Journals in Chemistry
subjects Artificial neural networks
Cartesian coordinates
Conformal mapping
Curves
Errors
Fluid dynamics
Fluid flow
Machine learning
Neural networks
Object recognition
Reconstruction
Reynolds number
Sensors
Training
Two dimensional flow
Vorticity
title Deep learning fluid flow reconstruction around arbitrary two-dimensional objects from sparse sensors using conformal mappings
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-08T03%3A04%3A42IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_scita&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Deep%20learning%20fluid%20flow%20reconstruction%20around%20arbitrary%20two-dimensional%20objects%20from%20sparse%20sensors%20using%20conformal%20mappings&rft.jtitle=AIP%20advances&rft.au=%C3%96zbay,%20Ali%20Girayhan&rft.date=2022-04-01&rft.volume=12&rft.issue=4&rft.spage=045126&rft.epage=045126-22&rft.pages=045126-045126-22&rft.issn=2158-3226&rft.eissn=2158-3226&rft.coden=AAIDBI&rft_id=info:doi/10.1063/5.0087488&rft_dat=%3Cproquest_scita%3E2655283453%3C/proquest_scita%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2655283453&rft_id=info:pmid/&rft_doaj_id=oai_doaj_org_article_d9c85f3166724668915780ee4bc769bd&rfr_iscdi=true