Hybrid FEM-NN models: Combining artificial neural networks with the finite element method
•An approach for training neural networks combined with partial differential equations.•Combines a finite element framework with a machine learning library.•Learns missing physics in the partial differential equation. We present a methodology combining neural networks with physical principle constra...
Saved in:
| Published in | Journal of computational physics Vol. 446; p. 110651 |
|---|---|
| Main Authors | , , |
| Format | Journal Article |
| Language | English |
| Published |
Cambridge
Elsevier Inc
01.12.2021
Elsevier Science Ltd |
| Subjects | |
| Online Access | Get full text |
| ISSN | 0021-9991 1090-2716 |
| DOI | 10.1016/j.jcp.2021.110651 |
Cover
| Abstract | •An approach for training neural networks combined with partial differential equations.•Combines a finite element framework with a machine learning library.•Learns missing physics in the partial differential equation.
We present a methodology combining neural networks with physical principle constraints in the form of partial differential equations (PDEs). The approach allows to train neural networks while respecting the PDEs as a strong constraint in the optimisation as apposed to making them part of the loss function. The resulting models are discretised in space by the finite element method (FEM). The method applies to both stationary and transient as well as linear/nonlinear PDEs. We describe implementation of the approach as an extension of the existing FEM framework FEniCS and its algorithmic differentiation tool dolfin-adjoint. Through series of examples we demonstrate capabilities of the approach to recover coefficients and missing PDE operators from observations. Further, the proposed method is compared with alternative methodologies, namely, physics informed neural networks and standard PDE-constrained optimisation. Finally, we demonstrate the method on a complex cardiac cell model problem using deep neural networks. |
|---|---|
| AbstractList | We present a methodology combining neural networks with physical principle constraints in the form of partial differential equations (PDEs). The approach allows to train neural networks while respecting the PDEs as a strong constraint in the optimisation as apposed to making them part of the loss function. The resulting models are discretised in space by the finite element method (FEM). The method applies to both stationary and transient as well as linear/nonlinear PDEs. We describe implementation of the approach as an extension of the existing FEM framework FEniCS and its algorithmic differentiation tool dolfin-adjoint. Through series of examples we demonstrate capabilities of the approach to recover coefficients and missing PDE operators from observations. Further, the proposed method is compared with alternative methodologies, namely, physics informed neural networks and standard PDE-constrained optimisation. Finally, we demonstrate the method on a complex cardiac cell model problem using deep neural networks. •An approach for training neural networks combined with partial differential equations.•Combines a finite element framework with a machine learning library.•Learns missing physics in the partial differential equation. We present a methodology combining neural networks with physical principle constraints in the form of partial differential equations (PDEs). The approach allows to train neural networks while respecting the PDEs as a strong constraint in the optimisation as apposed to making them part of the loss function. The resulting models are discretised in space by the finite element method (FEM). The method applies to both stationary and transient as well as linear/nonlinear PDEs. We describe implementation of the approach as an extension of the existing FEM framework FEniCS and its algorithmic differentiation tool dolfin-adjoint. Through series of examples we demonstrate capabilities of the approach to recover coefficients and missing PDE operators from observations. Further, the proposed method is compared with alternative methodologies, namely, physics informed neural networks and standard PDE-constrained optimisation. Finally, we demonstrate the method on a complex cardiac cell model problem using deep neural networks. |
| ArticleNumber | 110651 |
| Author | Funke, Simon W. Kuchta, Miroslav Mitusch, Sebastian K. |
| Author_xml | – sequence: 1 givenname: Sebastian K. orcidid: 0000-0002-8888-6568 surname: Mitusch fullname: Mitusch, Sebastian K. email: sebastkm@simula.no – sequence: 2 givenname: Simon W. orcidid: 0000-0003-4709-8415 surname: Funke fullname: Funke, Simon W. email: simon@simula.no – sequence: 3 givenname: Miroslav surname: Kuchta fullname: Kuchta, Miroslav email: miroslav@simula.no |
| BookMark | eNp9kLFu2zAQhokiBeK4fYBsBDrL4ckWRaZTYcRJANdd2qETQZPHmopEuiSdwG8fOe7UIdM_3P_d4b4rchFiQEKugc2AAb_pZp3Zz2pWwwyA8QY-kAkwyaq6BX5BJmycVFJKuCRXOXeMMdEsxIT8fjhuk7d0dfe92mzoEC32-ZYu47D1wYc_VKfinTde9zTgIb1FeYnpKdMXX3a07JC6sVqQYo8DhkIHLLtoP5GPTvcZP__LKfm1uvu5fKjWP-4fl9_WlZlzUSoDuhUcxNY2ujEtt8wtuNXbOQddAzdaCOmkY07AQtdNazm4mjEuec2F4M18Sr6c9-5T_HvAXFQXDymMJ1XdiEby8Ws5ttpzy6SYc0KnjC-6-BhK0r5XwNTJo-rU6FGdPKqzx5GE_8h98oNOx3eZr2dmlInPHpPKxmMwaH1CU5SN_h36FXnGi7I |
| CitedBy_id | crossref_primary_10_1007_s10462_024_10784_5 crossref_primary_10_1016_j_cma_2023_116692 crossref_primary_10_1016_j_mtcomm_2024_111278 crossref_primary_10_3390_math12111617 crossref_primary_10_1016_j_cma_2024_116890 crossref_primary_10_1016_j_neucom_2024_128936 crossref_primary_10_1016_j_cmpb_2023_107464 crossref_primary_10_1016_j_eml_2024_102157 crossref_primary_10_1038_s43017_023_00450_9 crossref_primary_10_1029_2024WR037318 crossref_primary_10_1016_j_physd_2024_134152 crossref_primary_10_1007_s10462_024_10836_w crossref_primary_10_1038_s44172_024_00303_3 crossref_primary_10_1016_j_jcp_2024_112791 crossref_primary_10_1063_5_0245547 crossref_primary_10_1137_23M1566935 crossref_primary_10_1007_s00419_024_02601_w crossref_primary_10_1038_s41598_022_16946_1 crossref_primary_10_1002_nme_70015 crossref_primary_10_1016_j_eml_2023_102103 crossref_primary_10_1002_celc_202300712 crossref_primary_10_1109_ACCESS_2024_3465572 crossref_primary_10_1016_j_compchemeng_2022_107898 crossref_primary_10_1016_j_ijnonlinmec_2024_104988 crossref_primary_10_1016_j_physd_2024_134362 crossref_primary_10_1016_j_cma_2025_117748 crossref_primary_10_1098_rsta_2022_0406 crossref_primary_10_1016_j_cma_2025_117889 crossref_primary_10_1088_1572_9494_accb8d crossref_primary_10_1103_PhysRevFluids_9_013901 crossref_primary_10_1109_TDMR_2024_3352001 crossref_primary_10_1093_imanum_drad073 crossref_primary_10_1016_j_jterra_2024_101003 crossref_primary_10_1016_j_tws_2024_111726 crossref_primary_10_1080_0952813X_2023_2242356 crossref_primary_10_1016_j_eml_2022_101645 crossref_primary_10_1016_j_enganabound_2023_08_020 crossref_primary_10_1016_j_engappai_2024_108010 crossref_primary_10_1016_j_jcp_2022_111768 crossref_primary_10_3390_math11092016 crossref_primary_10_1016_j_neucom_2025_129657 crossref_primary_10_3390_polym14235290 crossref_primary_10_1016_j_cma_2023_116505 crossref_primary_10_1016_j_soildyn_2024_109186 crossref_primary_10_1002_htj_23086 crossref_primary_10_1093_ptep_ptad131 crossref_primary_10_1140_epjp_s13360_024_05866_w crossref_primary_10_1016_j_cma_2024_116940 crossref_primary_10_1016_j_cmpb_2023_107616 crossref_primary_10_1016_j_ijsolstr_2024_112819 crossref_primary_10_1007_s00366_024_01955_7 crossref_primary_10_1016_j_engappai_2023_107324 crossref_primary_10_1016_j_engappai_2024_109128 crossref_primary_10_1016_j_jocs_2025_102533 crossref_primary_10_1038_s42005_024_01521_z crossref_primary_10_3934_math_20241683 crossref_primary_10_1016_j_jocs_2023_102136 crossref_primary_10_3390_machines11111013 crossref_primary_10_1016_j_compstruc_2025_107698 crossref_primary_10_1016_j_cma_2022_115766 crossref_primary_10_1007_s00366_024_02033_8 |
| Cites_doi | 10.1145/279232.279236 10.1007/BF01589116 10.1016/j.jcp.2019.109136 10.1113/jphysiol.1962.sp006849 10.1016/j.cma.2020.113028 10.1016/j.jcp.2018.10.045 10.1038/s41592-019-0686-2 10.1145/2566630 10.1016/j.jcp.2020.109491 10.1073/pnas.1517384113 10.1016/S0898-1221(99)00248-5 10.1146/annurev-fluid-010518-040547 10.1023/B:BITN.0000046811.70614.38 10.1113/jphysiol.1952.sp004764 10.1145/1731022.1731030 10.1016/j.jcp.2019.108925 10.1038/s41598-020-66042-5 10.1137/17M1144532 10.1137/0721052 |
| ContentType | Journal Article |
| Copyright | 2021 The Authors Copyright Elsevier Science Ltd. Dec 1, 2021 |
| Copyright_xml | – notice: 2021 The Authors – notice: Copyright Elsevier Science Ltd. Dec 1, 2021 |
| DBID | 6I. AAFTH AAYXX CITATION 7SC 7SP 7U5 8FD JQ2 L7M L~C L~D |
| DOI | 10.1016/j.jcp.2021.110651 |
| DatabaseName | ScienceDirect Open Access Titles Elsevier:ScienceDirect:Open Access CrossRef Computer and Information Systems Abstracts Electronics & Communications Abstracts Solid State and Superconductivity Abstracts Technology Research Database ProQuest Computer Science Collection Advanced Technologies Database with Aerospace Computer and Information Systems Abstracts Academic Computer and Information Systems Abstracts Professional |
| DatabaseTitle | CrossRef Technology Research Database Computer and Information Systems Abstracts – Academic Electronics & Communications Abstracts ProQuest Computer Science Collection Computer and Information Systems Abstracts Solid State and Superconductivity Abstracts Advanced Technologies Database with Aerospace Computer and Information Systems Abstracts Professional |
| DatabaseTitleList | Technology Research Database |
| DeliveryMethod | fulltext_linktorsrc |
| Discipline | Applied Sciences |
| EISSN | 1090-2716 |
| ExternalDocumentID | 10_1016_j_jcp_2021_110651 S0021999121005465 |
| GroupedDBID | --K --M -~X .~1 0R~ 1B1 1RT 1~. 1~5 4.4 457 4G. 5GY 5VS 6I. 6OB 7-5 71M 8P~ 9JN AABNK AACTN AAEDT AAEDW AAFTH AAIAV AAIKJ AAKOC AALRI AAOAW AAQFI AAXUO AAYFN ABBOA ABFRF ABJNI ABMAC ABNEU ABYKQ ACBEA ACDAQ ACFVG ACGFO ACGFS ACNCT ACRLP ACZNC ADBBV ADEZE AEBSH AEFWE AEKER AENEX AFKWA AFTJW AGHFR AGUBO AGYEJ AHHHB AHZHX AIALX AIEXJ AIKHN AITUG AIVDX AJOXV ALMA_UNASSIGNED_HOLDINGS AMFUW AMRAJ AOUOD AXJTR BKOJK BLXMC CS3 DM4 DU5 EBS EFBJH EFLBG EO8 EO9 EP2 EP3 F5P FDB FEDTE FIRID FNPLU FYGXN G-Q GBLVA GBOLZ HLZ HVGLF IHE J1W K-O KOM LG5 LX9 LZ4 M37 M41 MO0 N9A O-L O9- OAUVE OGIMB OZT P-8 P-9 P2P PC. Q38 RNS ROL RPZ SDF SDG SDP SES SPC SPCBC SPD SSQ SSV SSZ T5K TN5 UPT YQT ZMT ZU3 ~02 ~G- 29K 6TJ 8WZ A6W AAQXK AATTM AAXKI AAYWO AAYXX ABFNM ABWVN ABXDB ACNNM ACRPL ACVFH ADCNI ADFGL ADIYS ADJOM ADMUD ADNMO AEIPS AEUPX AFFNX AFJKZ AFPUW AFXIZ AGCQF AGQPQ AGRNS AIGII AIIUN AKBMS AKRWK AKYEP ANKPU APXCP ASPBG AVWKF AZFZN BBWZM BNPGV CAG CITATION COF D-I EJD FGOYB G-2 HME HMV HZ~ NDZJH R2- RIG SBC SEW SHN SPG SSH T9H UQL WUQ ZY4 7SC 7SP 7U5 8FD EFKBS JQ2 L7M L~C L~D |
| ID | FETCH-LOGICAL-c368t-c1a78618bd5a5c76d0f46dab361a216ca889f9f0f814a257d61f2006962688653 |
| IEDL.DBID | AIKHN |
| ISSN | 0021-9991 |
| IngestDate | Fri Jul 25 06:13:30 EDT 2025 Thu Apr 24 23:00:07 EDT 2025 Tue Jul 01 01:54:52 EDT 2025 Fri Feb 23 02:43:09 EST 2024 |
| IsDoiOpenAccess | true |
| IsOpenAccess | true |
| IsPeerReviewed | true |
| IsScholarly | true |
| Keywords | Finite element method Data-driven scientific computing Partial differential equations Learning unknown physics Machine learning |
| Language | English |
| License | This is an open access article under the CC BY-NC-ND license. |
| LinkModel | DirectLink |
| MergedId | FETCHMERGED-LOGICAL-c368t-c1a78618bd5a5c76d0f46dab361a216ca889f9f0f814a257d61f2006962688653 |
| Notes | ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14 |
| ORCID | 0000-0003-4709-8415 0000-0002-8888-6568 |
| OpenAccessLink | https://www.sciencedirect.com/science/article/pii/S0021999121005465 |
| PQID | 2585969999 |
| PQPubID | 2047462 |
| ParticipantIDs | proquest_journals_2585969999 crossref_citationtrail_10_1016_j_jcp_2021_110651 crossref_primary_10_1016_j_jcp_2021_110651 elsevier_sciencedirect_doi_10_1016_j_jcp_2021_110651 |
| ProviderPackageCode | CITATION AAYXX |
| PublicationCentury | 2000 |
| PublicationDate | 2021-12-01 2021-12-00 20211201 |
| PublicationDateYYYYMMDD | 2021-12-01 |
| PublicationDate_xml | – month: 12 year: 2021 text: 2021-12-01 day: 01 |
| PublicationDecade | 2020 |
| PublicationPlace | Cambridge |
| PublicationPlace_xml | – name: Cambridge |
| PublicationTitle | Journal of computational physics |
| PublicationYear | 2021 |
| Publisher | Elsevier Inc Elsevier Science Ltd |
| Publisher_xml | – name: Elsevier Inc – name: Elsevier Science Ltd |
| References | Liu, Nocedal (br0300) 1989; 45 Ruthotto, Haber (br0040) 2019 Naumann (br0250) 2012 Li, Kovachki, Azizzadenesheli, Liu, Bhattacharya, Stuart, Anandkumar (br0060) Kværnø (br0350) 2004; 44 Raissi, Perdikaris, Karniadakis (br0070) 2019; 378 Wang, Yu, Perdikaris (br0110) Rackauckas, Ma, Martensen, Warner, Zubov, Supekar, Skinner, Ramadhan (br0180) Glowinski (br0230) 2003; vol. 9 Sundnes, Lines, Cai, Nielsen, Mardal, Tveito (br0430) 2007 Duraisamy, Iaccarino, Xiao (br0210) 2019; 51 Zhu, Byrd, Lu, Nocedal (br0420) 1997; 23 Tartakovsky, Marrero, Perdikaris, Tartakovsky, Barajas-Solano (br0080) Marcus (br0010) Shen (br0370) 1999; 38 Fan, Xu, Pathak, Darve (br0160) Valnes, Mitusch, Ringstad, Eide, Funke, Mardal (br0340) 2020; 10 Paszke, Gross, Massa, Lerer, Bradbury, Chanan, Killeen, Lin, Gimelshein, Antiga, Desmaison, Kopf, Yang, DeVito, Raison, Tejani, Chilamkurthy, Steiner, Fang, Bai, Chintala (br0200) 2019; vol. 32 Xu, Darve (br0150) Kharazmi, Zhang, Karniadakis (br0240) Brunton, Proctor, Kutz (br0220) 2016; 113 Hodgkin, Huxley (br0450) 1952; 117 Virtanen, Gommers, Oliphant, Haberland, Reddy, Cournapeau, Burovski, Peterson, Weckesser, Bright, van der Walt, Brett, Wilson, Millman, Mayorov, Nelson, Jones, Kern, Larson, Carey, Polat, Feng, Moore, VanderPlas, Laxalde, Perktold, Cimrman, Henriksen, Quintero, Harris, Archibald, Ribeiro, Pedregosa, van Mulbregt (br0310) 2020; 17 Ayed, de Bézenac, Pajot, Brajard, Gallinari (br0470) Xu, Tartakovsky, Burghardt, Darve (br0170) Mishra, Molinaro (br0120) Logg, Wells (br0190) 2010; 37 Homolya, Kirby, Ham (br0290) Huang, Xu, Farhat, Darve (br0140) 2020; 416 Nash (br0320) 1984; 21 Berg, Nyström (br0130) Jagtap, Karniadakis (br0390) 2020 Wang, Teng, Perdikaris (br0400) Jagtap, Kawaguchi, Karniadakis (br0410) 2020; 404 Shin, Darbon, Karniadakis (br0100) Skare (br0360) 2012 Mitusch, Funke, Dokken (br0280) 2019; 4 Alnæs, Logg, Ølgaard, Rognes, Wells (br0270) 2014; 40 Long, Lu, Ma, Dong (br0030) 2018 Lu, Jin, Karniadakis (br0050) Ulbrich (br0260) 2009 Löhner, Antil (br0330) Raissi (br0090) 2018; 19 Farrell, Hake, Funke, Rognes (br0440) 2019; 41 Long, Lu, Dong (br0020) 2019; 399 Noble (br0460) 1962; 160 Jagtap, Kharazmi, Karniadakis (br0380) 2020; 365 Naumann (10.1016/j.jcp.2021.110651_br0250) 2012 Logg (10.1016/j.jcp.2021.110651_br0190) 2010; 37 Liu (10.1016/j.jcp.2021.110651_br0300) 1989; 45 Jagtap (10.1016/j.jcp.2021.110651_br0410) 2020; 404 Mitusch (10.1016/j.jcp.2021.110651_br0280) 2019; 4 Farrell (10.1016/j.jcp.2021.110651_br0440) 2019; 41 Marcus (10.1016/j.jcp.2021.110651_br0010) Sundnes (10.1016/j.jcp.2021.110651_br0430) 2007 Homolya (10.1016/j.jcp.2021.110651_br0290) Ruthotto (10.1016/j.jcp.2021.110651_br0040) 2019 Rackauckas (10.1016/j.jcp.2021.110651_br0180) Brunton (10.1016/j.jcp.2021.110651_br0220) 2016; 113 Skare (10.1016/j.jcp.2021.110651_br0360) 2012 Huang (10.1016/j.jcp.2021.110651_br0140) 2020; 416 Jagtap (10.1016/j.jcp.2021.110651_br0390) 2020 Löhner (10.1016/j.jcp.2021.110651_br0330) Duraisamy (10.1016/j.jcp.2021.110651_br0210) 2019; 51 Hodgkin (10.1016/j.jcp.2021.110651_br0450) 1952; 117 Raissi (10.1016/j.jcp.2021.110651_br0090) 2018; 19 Shin (10.1016/j.jcp.2021.110651_br0100) Glowinski (10.1016/j.jcp.2021.110651_br0230) 2003; vol. 9 Ayed (10.1016/j.jcp.2021.110651_br0470) Ulbrich (10.1016/j.jcp.2021.110651_br0260) 2009 Wang (10.1016/j.jcp.2021.110651_br0400) Mishra (10.1016/j.jcp.2021.110651_br0120) Fan (10.1016/j.jcp.2021.110651_br0160) Kværnø (10.1016/j.jcp.2021.110651_br0350) 2004; 44 Shen (10.1016/j.jcp.2021.110651_br0370) 1999; 38 Long (10.1016/j.jcp.2021.110651_br0030) 2018 Valnes (10.1016/j.jcp.2021.110651_br0340) 2020; 10 Lu (10.1016/j.jcp.2021.110651_br0050) Long (10.1016/j.jcp.2021.110651_br0020) 2019; 399 Paszke (10.1016/j.jcp.2021.110651_br0200) 2019; vol. 32 Tartakovsky (10.1016/j.jcp.2021.110651_br0080) Wang (10.1016/j.jcp.2021.110651_br0110) Zhu (10.1016/j.jcp.2021.110651_br0420) 1997; 23 Alnæs (10.1016/j.jcp.2021.110651_br0270) 2014; 40 Xu (10.1016/j.jcp.2021.110651_br0150) Virtanen (10.1016/j.jcp.2021.110651_br0310) 2020; 17 Jagtap (10.1016/j.jcp.2021.110651_br0380) 2020; 365 Berg (10.1016/j.jcp.2021.110651_br0130) Xu (10.1016/j.jcp.2021.110651_br0170) Kharazmi (10.1016/j.jcp.2021.110651_br0240) Noble (10.1016/j.jcp.2021.110651_br0460) 1962; 160 Li (10.1016/j.jcp.2021.110651_br0060) Nash (10.1016/j.jcp.2021.110651_br0320) 1984; 21 Raissi (10.1016/j.jcp.2021.110651_br0070) 2019; 378 |
| References_xml | – volume: 19 start-page: 1 year: 2018 end-page: 24 ident: br0090 article-title: Deep hidden physics models: deep learning of nonlinear partial differential equations publication-title: J. Mach. Learn. Res. – volume: vol. 9 start-page: 3 year: 2003 end-page: 1176 ident: br0230 article-title: Finite element methods for incompressible viscous flow publication-title: Numerical Methods for Fluids (Part 3) – volume: 4 year: 2019 ident: br0280 article-title: Dolfin-adjoint 2018.1: automated adjoints for FEniCS and Firedrake publication-title: J. Open Sour. Softw. – volume: 45 start-page: 503 year: 1989 end-page: 528 ident: br0300 article-title: On the limited memory bfgs method for large scale optimization publication-title: Math. Program. – volume: 113 start-page: 3932 year: 2016 end-page: 3937 ident: br0220 article-title: Discovering governing equations from data by sparse identification of nonlinear dynamical systems publication-title: Proc. Natl. Acad. Sci. – volume: 399 year: 2019 ident: br0020 article-title: PDE-Net 2.0: learning PDEs from data with a numeric-symbolic hybrid deep network publication-title: J. Comput. Phys. – start-page: 2002 year: 2020 end-page: 2041 ident: br0390 article-title: Extended physics-informed neural networks (XPINNs): a generalized space-time domain decomposition based deep learning framework for nonlinear partial differential equations publication-title: Commun. Comput. Phys. – year: 2007 ident: br0430 article-title: Computing the Electrical Activity in the Heart publication-title: Monographs in Computational Science and Engineering – ident: br0290 article-title: Exposing and exploiting structure: optimal code generation for high-order finite element methods – volume: 38 start-page: 173 year: 1999 end-page: 188 ident: br0370 article-title: A numerical study of inverse heat conduction problems publication-title: Comput. Math. Appl. – volume: 365 year: 2020 ident: br0380 article-title: Conservative physics-informed neural networks on discrete domains for conservation laws: applications to forward and inverse problems publication-title: Comput. Methods Appl. Mech. Eng. – volume: vol. 32 start-page: 8024 year: 2019 end-page: 8035 ident: br0200 article-title: Pytorch: an imperative style, high-performance deep learning library publication-title: Advances in Neural Information Processing Systems – ident: br0400 article-title: Understanding and mitigating gradient pathologies in physics-informed neural networks – volume: 44 start-page: 489 year: 2004 end-page: 502 ident: br0350 article-title: Singly diagonally implicit Runge-Kutta methods with an explicit first stage publication-title: BIT Numer. Math. – ident: br0150 article-title: Adcme: learning spatially-varying physical fields using deep neural networks – ident: br0130 article-title: Neural network augmented inverse problems for PDEs – ident: br0470 article-title: Learning dynamical systems from partial observations – volume: 17 start-page: 261 year: 2020 end-page: 272 ident: br0310 article-title: SciPy 1.0 contributors, SciPy 1.0: fundamental algorithms for scientific computing in Python publication-title: Nat. Methods – ident: br0060 article-title: Fourier neural operator for parametric partial differential equations – volume: 378 start-page: 686 year: 2019 end-page: 707 ident: br0070 article-title: Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations publication-title: J. Comput. Phys. – volume: 416 year: 2020 ident: br0140 article-title: Learning constitutive relations from indirect observations using deep neural networks publication-title: J. Comput. Phys. – volume: 23 start-page: 550 year: 1997 end-page: 560 ident: br0420 article-title: Algorithm 778: L-BFGS-B: Fortran subroutines for large-scale bound-constrained optimization publication-title: ACM Trans. Math. Softw. – volume: 117 start-page: 500 year: 1952 end-page: 544 ident: br0450 article-title: A quantitative description of membrane current and its application to conduction and excitation in nerve publication-title: J. Physiol. – volume: 404 year: 2020 ident: br0410 article-title: Adaptive activation functions accelerate convergence in deep and physics-informed neural networks publication-title: J. Comput. Phys. – ident: br0010 article-title: Deep learning: a critical appraisal – ident: br0050 article-title: Deeponet: learning nonlinear operators for identifying differential equations based on the universal approximation theorem of operators – volume: 37 start-page: 20:1 year: 2010 end-page: 20:28 ident: br0190 article-title: DOLFIN: automated finite element computing publication-title: ACM Trans. Math. Softw. – volume: 160 start-page: 317 year: 1962 end-page: 352 ident: br0460 article-title: A modification of the Hodgkin-Huxley equations applicable to Purkinje fibre action and pacemaker potentials publication-title: J. Physiol. – ident: br0100 article-title: On the convergence and generalization of physics informed neural networks – volume: 21 start-page: 770 year: 1984 end-page: 788 ident: br0320 article-title: Newton-type minimization via the Lanczos method publication-title: SIAM J. Numer. Anal. – ident: br0160 article-title: Solving inverse problems in steady state Navier-Stokes equations using deep neural networks – year: 2012 ident: br0360 article-title: Gryphon-a module for time integration of partial differential equations in fenics – volume: 41 start-page: C219 year: 2019 end-page: C244 ident: br0440 article-title: Automated adjoints of coupled PDE-ODE systems publication-title: SIAM J. Sci. Comput. – ident: br0330 article-title: Revisiting Calderon's problem – start-page: 1 year: 2019 end-page: 13 ident: br0040 article-title: Deep neural networks motivated by partial differential equations publication-title: J. Math. Imaging Vis. – ident: br0120 article-title: Estimates on the generalization error of physics informed neural networks (pinns) for approximating pdes – ident: br0170 article-title: Inverse modeling of viscoelasticity materials using physics constrained learning – volume: 40 start-page: 9:1 year: 2014 end-page: 9:37 ident: br0270 article-title: Unified form language: a domain-specific language for weak formulations of partial differential equations publication-title: ACM Trans. Math. Softw. – ident: br0110 article-title: When and why pinns fail to train: a neural tangent kernel perspective – ident: br0080 article-title: Learning parameters and constitutive relationships with physics informed deep neural networks – volume: 51 start-page: 357 year: 2019 end-page: 377 ident: br0210 article-title: Turbulence modeling in the age of data publication-title: Annu. Rev. Fluid Mech. – start-page: 3208 year: 2018 end-page: 3216 ident: br0030 article-title: PDE-Net: learning PDEs from data publication-title: International Conference on Machine Learning – volume: 10 start-page: 1 year: 2020 end-page: 12 ident: br0340 article-title: Apparent diffusion coefficient estimates based on 24 hours tracer movement support glymphatic transport in human cerebral cortex publication-title: Sci. Rep. – start-page: 1 year: 2009 end-page: 95 ident: br0260 article-title: Analytical Background and Optimality Theory – ident: br0180 article-title: Universal differential equations for scientific machine learning – ident: br0240 article-title: Variational physics-informed neural networks for solving partial differential equations – year: 2012 ident: br0250 article-title: The Art of Differentiating Computer Programs: An Introduction to Algorithmic Differentiation – volume: 23 start-page: 550 issue: 4 year: 1997 ident: 10.1016/j.jcp.2021.110651_br0420 article-title: Algorithm 778: L-BFGS-B: Fortran subroutines for large-scale bound-constrained optimization publication-title: ACM Trans. Math. Softw. doi: 10.1145/279232.279236 – year: 2012 ident: 10.1016/j.jcp.2021.110651_br0360 – ident: 10.1016/j.jcp.2021.110651_br0180 – ident: 10.1016/j.jcp.2021.110651_br0060 – ident: 10.1016/j.jcp.2021.110651_br0110 – volume: 45 start-page: 503 issue: 1 year: 1989 ident: 10.1016/j.jcp.2021.110651_br0300 article-title: On the limited memory bfgs method for large scale optimization publication-title: Math. Program. doi: 10.1007/BF01589116 – ident: 10.1016/j.jcp.2021.110651_br0080 – volume: vol. 9 start-page: 3 year: 2003 ident: 10.1016/j.jcp.2021.110651_br0230 article-title: Finite element methods for incompressible viscous flow – ident: 10.1016/j.jcp.2021.110651_br0470 – volume: 404 year: 2020 ident: 10.1016/j.jcp.2021.110651_br0410 article-title: Adaptive activation functions accelerate convergence in deep and physics-informed neural networks publication-title: J. Comput. Phys. doi: 10.1016/j.jcp.2019.109136 – ident: 10.1016/j.jcp.2021.110651_br0100 – ident: 10.1016/j.jcp.2021.110651_br0130 – start-page: 2002 year: 2020 ident: 10.1016/j.jcp.2021.110651_br0390 article-title: Extended physics-informed neural networks (XPINNs): a generalized space-time domain decomposition based deep learning framework for nonlinear partial differential equations publication-title: Commun. Comput. Phys. – volume: 160 start-page: 317 issue: 2 year: 1962 ident: 10.1016/j.jcp.2021.110651_br0460 article-title: A modification of the Hodgkin-Huxley equations applicable to Purkinje fibre action and pacemaker potentials publication-title: J. Physiol. doi: 10.1113/jphysiol.1962.sp006849 – volume: 365 year: 2020 ident: 10.1016/j.jcp.2021.110651_br0380 article-title: Conservative physics-informed neural networks on discrete domains for conservation laws: applications to forward and inverse problems publication-title: Comput. Methods Appl. Mech. Eng. doi: 10.1016/j.cma.2020.113028 – volume: 378 start-page: 686 year: 2019 ident: 10.1016/j.jcp.2021.110651_br0070 article-title: Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations publication-title: J. Comput. Phys. doi: 10.1016/j.jcp.2018.10.045 – volume: 19 start-page: 1 issue: 25 year: 2018 ident: 10.1016/j.jcp.2021.110651_br0090 article-title: Deep hidden physics models: deep learning of nonlinear partial differential equations publication-title: J. Mach. Learn. Res. – volume: 17 start-page: 261 year: 2020 ident: 10.1016/j.jcp.2021.110651_br0310 article-title: SciPy 1.0 contributors, SciPy 1.0: fundamental algorithms for scientific computing in Python publication-title: Nat. Methods doi: 10.1038/s41592-019-0686-2 – volume: 40 start-page: 9:1 issue: 2 year: 2014 ident: 10.1016/j.jcp.2021.110651_br0270 article-title: Unified form language: a domain-specific language for weak formulations of partial differential equations publication-title: ACM Trans. Math. Softw. doi: 10.1145/2566630 – start-page: 1 year: 2019 ident: 10.1016/j.jcp.2021.110651_br0040 article-title: Deep neural networks motivated by partial differential equations publication-title: J. Math. Imaging Vis. – volume: 416 year: 2020 ident: 10.1016/j.jcp.2021.110651_br0140 article-title: Learning constitutive relations from indirect observations using deep neural networks publication-title: J. Comput. Phys. doi: 10.1016/j.jcp.2020.109491 – volume: 113 start-page: 3932 issue: 15 year: 2016 ident: 10.1016/j.jcp.2021.110651_br0220 article-title: Discovering governing equations from data by sparse identification of nonlinear dynamical systems publication-title: Proc. Natl. Acad. Sci. doi: 10.1073/pnas.1517384113 – volume: 4 issue: 38 year: 2019 ident: 10.1016/j.jcp.2021.110651_br0280 article-title: Dolfin-adjoint 2018.1: automated adjoints for FEniCS and Firedrake publication-title: J. Open Sour. Softw. – year: 2007 ident: 10.1016/j.jcp.2021.110651_br0430 article-title: Computing the Electrical Activity in the Heart – volume: 38 start-page: 173 issue: 7–8 year: 1999 ident: 10.1016/j.jcp.2021.110651_br0370 article-title: A numerical study of inverse heat conduction problems publication-title: Comput. Math. Appl. doi: 10.1016/S0898-1221(99)00248-5 – volume: 51 start-page: 357 issue: 1 year: 2019 ident: 10.1016/j.jcp.2021.110651_br0210 article-title: Turbulence modeling in the age of data publication-title: Annu. Rev. Fluid Mech. doi: 10.1146/annurev-fluid-010518-040547 – start-page: 1 year: 2009 ident: 10.1016/j.jcp.2021.110651_br0260 – volume: 44 start-page: 489 issue: 3 year: 2004 ident: 10.1016/j.jcp.2021.110651_br0350 article-title: Singly diagonally implicit Runge-Kutta methods with an explicit first stage publication-title: BIT Numer. Math. doi: 10.1023/B:BITN.0000046811.70614.38 – volume: 117 start-page: 500 issue: 4 year: 1952 ident: 10.1016/j.jcp.2021.110651_br0450 article-title: A quantitative description of membrane current and its application to conduction and excitation in nerve publication-title: J. Physiol. doi: 10.1113/jphysiol.1952.sp004764 – volume: 37 start-page: 20:1 issue: 2 year: 2010 ident: 10.1016/j.jcp.2021.110651_br0190 article-title: DOLFIN: automated finite element computing publication-title: ACM Trans. Math. Softw. doi: 10.1145/1731022.1731030 – ident: 10.1016/j.jcp.2021.110651_br0170 – ident: 10.1016/j.jcp.2021.110651_br0050 – volume: vol. 32 start-page: 8024 year: 2019 ident: 10.1016/j.jcp.2021.110651_br0200 article-title: Pytorch: an imperative style, high-performance deep learning library – ident: 10.1016/j.jcp.2021.110651_br0400 – volume: 399 year: 2019 ident: 10.1016/j.jcp.2021.110651_br0020 article-title: PDE-Net 2.0: learning PDEs from data with a numeric-symbolic hybrid deep network publication-title: J. Comput. Phys. doi: 10.1016/j.jcp.2019.108925 – volume: 10 start-page: 1 issue: 1 year: 2020 ident: 10.1016/j.jcp.2021.110651_br0340 article-title: Apparent diffusion coefficient estimates based on 24 hours tracer movement support glymphatic transport in human cerebral cortex publication-title: Sci. Rep. doi: 10.1038/s41598-020-66042-5 – volume: 41 start-page: C219 issue: 3 year: 2019 ident: 10.1016/j.jcp.2021.110651_br0440 article-title: Automated adjoints of coupled PDE-ODE systems publication-title: SIAM J. Sci. Comput. doi: 10.1137/17M1144532 – start-page: 3208 year: 2018 ident: 10.1016/j.jcp.2021.110651_br0030 article-title: PDE-Net: learning PDEs from data – ident: 10.1016/j.jcp.2021.110651_br0120 – ident: 10.1016/j.jcp.2021.110651_br0290 – volume: 21 start-page: 770 issue: 4 year: 1984 ident: 10.1016/j.jcp.2021.110651_br0320 article-title: Newton-type minimization via the Lanczos method publication-title: SIAM J. Numer. Anal. doi: 10.1137/0721052 – ident: 10.1016/j.jcp.2021.110651_br0010 – ident: 10.1016/j.jcp.2021.110651_br0240 – ident: 10.1016/j.jcp.2021.110651_br0150 – ident: 10.1016/j.jcp.2021.110651_br0160 – year: 2012 ident: 10.1016/j.jcp.2021.110651_br0250 – ident: 10.1016/j.jcp.2021.110651_br0330 |
| SSID | ssj0008548 |
| Score | 2.6123803 |
| Snippet | •An approach for training neural networks combined with partial differential equations.•Combines a finite element framework with a machine learning... We present a methodology combining neural networks with physical principle constraints in the form of partial differential equations (PDEs). The approach... |
| SourceID | proquest crossref elsevier |
| SourceType | Aggregation Database Enrichment Source Index Database Publisher |
| StartPage | 110651 |
| SubjectTerms | Artificial neural networks Computational physics Constraints Data-driven scientific computing Finite element analysis Finite element method Learning unknown physics Machine learning Mathematical models Neural networks Operators (mathematics) Optimization Partial differential equations |
| Title | Hybrid FEM-NN models: Combining artificial neural networks with the finite element method |
| URI | https://dx.doi.org/10.1016/j.jcp.2021.110651 https://www.proquest.com/docview/2585969999 |
| Volume | 446 |
| hasFullText | 1 |
| inHoldings | 1 |
| isFullTextHit | |
| isPrint | |
| link | http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwnV07T8MwED6VsrDwRjwK8sCEFIjjR202VFEVEJ1AgslynFgqQgXRMrDw2_HFTiWQYGCKEsVRcuf7cmd_dwdwXJaM8xJTtwR1GXe1zkqmbcaYqPI6hHK5xx3d27Ec3fPrB_HQgUGbC4O0yoT9EdMbtE5XzpI0z14nE8zxLTCHHitg5djSewmWC6al6MLyxdXNaLwAZCV4BGRkI4QB7eZmQ_N6cli1sqDIh5eC_vZ7-gHUzd9nuA6ryW0kF_HNNqBTTzdhLbmQJBnobAseRx-YgkWGl7fZeEyaPjezcxKsvmw6QRD8tlg0gmApy-bQEMFnBJdkSfAHiZ-gI0rqyCwnscn0NtwPL-8Goyx1T8gck2qeOWr7SlJVVsIK15dV7rmsbMkktQWVziqlvfZBF5TbYLiVpB7XF3QIcZSSgu1Ad_oyrXeBFDooOlcVqyrH-7m3zkol6txKX4dw0O9B3grNuFRaHDtcPJuWQ_ZkgpwNytlEOe_ByWLIa6yr8dfNvNWE-TY5TMD9v4b1Wq2ZZJkzU4T4SMswDfT-_556ACt4FiktPejO397rw-CYzMsjWDr9pEdp-n0B1ODgbw |
| linkProvider | Elsevier |
| linkToHtml | http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwnV07T8MwELZ4DLDwRjwKeGBCCo3jRx02VLUqj3ZqpTJZjhNLrVCpSBlY-O34YgcEEh2YIiV2lNzLd_Z3dwhdZhllLIPULU5MxEyRRhlNdUQpz-PChXKxhRPd_kD0Rux-zMcrqF3nwgCsMth-b9Mrax3uNAM1m_PJBHJ8E8ihhwpYMbT0XkXrjNMW4PquP75xHpIzb44Bi-CG10ebFchraqBmZUIADS84-Wtx-mWmq7Wnu4O2gtOIb_137aKVYraHtoMDiYN6lvvoqfcOCVi42-lHgwGuutyUN9jpfFb1gcDwZ75kBIZCltWlgoGXGDZksfMGsZ2AG4oLjyvHvsX0ARp1O8N2Lwq9EyJDhVxEhuiWFERmOdfctEQeWyZynVFBdEKE0VKmNrWOE4Rpp7a5IBZ2F1IX4EgpOD1Ea7OXWXGEcJI6Nscyp3luWCu22mgheRFrYQsXDNpjFNdEUyYUFof-Fs-qRpBNlaOzAjorT-djdPU1Ze6raiwbzGpOqB-ioZzVXzatUXNNBb0sVeKio1Q4MUhP_vfWC7TRG_Yf1ePd4OEUbcITD25poLXF61tx5lyURXZeieAntPDhOg |
| openUrl | ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Hybrid+FEM-NN+models%3A+Combining+artificial+neural+networks+with+the+finite+element+method&rft.jtitle=Journal+of+computational+physics&rft.au=Mitusch%2C+Sebastian+K&rft.au=Funke%2C+Simon+W&rft.au=Kuchta%2C+Miroslav&rft.date=2021-12-01&rft.pub=Elsevier+Science+Ltd&rft.issn=0021-9991&rft.eissn=1090-2716&rft.volume=446&rft.spage=1&rft_id=info:doi/10.1016%2Fj.jcp.2021.110651&rft.externalDBID=NO_FULL_TEXT |
| thumbnail_l | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=0021-9991&client=summon |
| thumbnail_m | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=0021-9991&client=summon |
| thumbnail_s | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=0021-9991&client=summon |