Domain decomposition based physics and equality constrained neural networks for the Helmholtz equation

i

This article has been updated. A Correction to this article was published on .

i

The Original Article was published on

i

The Original Article was published on

i

This article has been updated

i

This article was retracted on

i

This article has been updated

Abstract

Overlapping and non-overlapping domain decomposition methods are proposed for the physics and equality constrained neural networks (PECANN) to solve the Helmholtz equation. A coarse component is introduced using a small neural network defined on the whole domain with carefully designed loss functions. This coarse component supplies both function values and normal derivatives on the subdomain interfaces to the local solvers, ensuring the scalability of the algorithms; the number of outer iterations remains nearly constant as the number of subdomains increases. Numerical experiments conducted with a wide range of wavenumbers over both standard and enlarged domains demonstrate the efficiency and scalability of the proposed algorithms.

Keywords

PINN / PECANN / domain decomposition / overlapping / non-overlapping / coarse component / Helmholtz

CHANGE HISTORY

A Correction to this paper has been published:
https://doi.org

The Retraction Note was updated to reflect reception of statements of disagreement from all authors.

This article has been retracted.Please see the Retraction Notice for more detail:
https://doi.org

References

[1] I. M. Babuška, S. A. Sauter, Is the pollution effect of the FEM avoidable for the Helmholtz equation considering high wave numbers? SIAM J. Numer. Anal., 34 (1997), 2392-2423. https://doi.org/10.1137/S0036142994269186

[2] W. Chen, Y. Liu, X. Xu, A robust domain decomposition method for the Helmholtz equation with high wave number, ESAIM. Math. Model. Numer. Anal., 50 (2016), 921-944. https://doi.org/10.1051/m2an/2015058

[3] C. Farhat, A. Macedo, M. Lesoinne, A two-level domain decomposition method for the iterative solution of high-frequency exterior Helmholtz problems, Numer. Math., 85 (2000), 283-308. https://doi.org/10.1007/PL00005389

[4] M. J. Gander, F. Magoulès, F. Nataf, Optimized Schwarz methods without overlap for the Helmholtz equation, SIAM J. Sci. Comput., 24 (2002), 38-60. https://doi.org/10.1137/S1064827501387012

[5] C. Farhat, J. Li, An iterative domain decomposition method for the solution of a class of indefinite problems in computational structural dynamics, Appl. Numer. Math., 54 (2005), 150-166. https://doi.org/10.1016/j.apnum.2004.09.021

[6] M. J. Gander, L. Halpern, F. Magoulès, An optimized Schwarz method with two-sided Robin transmission conditions for the Helmholtz equation, Int. J. Numer. Methods Fluids, 55 (2007), 163-175. https://doi.org/10.1002/fld.1433

[7] J. Li, X. Tu, Convergence analysis of a balancing domain decomposition method for solving a class of indefinite linear systems, Numer. Linear Algebra Appl., 16 (2009), 745-773. https://doi.org/10.1002/nla.639

[8] M. J. Gander, H. Zhang, Optimized Schwarz methods with overlap for the Helmholtz equation, SIAM J. Sci. Comput., 38 (2016), A3195-A3219. https://doi.org/10.1137/15M1021659

[9] P. Lu, X. Xu, A robust multilevel preconditioner based on a domain decomposition method for the Helmholtz equation, J. Sci. Comput., 81 (2019), 291-311. https://doi.org/10.1007/s10915-019-01015-z

[10] Y. Liu, X. Xu, An optimized Schwarz method with relaxation for the Helmholtz equation: the negative impact of overlap, ESAIM. Math. Model. Numer. Anal., 53 (2019), 249-268. https://doi.org/10.1051/m2an/2018061

[11] K. Hornik, M. Stinchcombe, H. White, Multilayer feedforward networks are universal approximators, Neural Networks, 2 (1989), 359-366. https://doi.org/10.1016/0893-6080(89)90020-8

[12] M. Raissi, P. Perdikaris, G. E. Karniadakis, Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations, J. Comput. Phys., 378 (2019), 686-707. https://doi.org/10.1016/j.jcp.2018.10.045

[13] S. Cai, Z. Mao, Z. Wang, M. Yin, G. Karniadakis, Physics-informed neural networks (PINNs) for fluid mechanics: a review, Acta Mech. Sin., 37 (2021), 1727-1738. https://doi.org/10.1007/s10409-021-01148-1

[14] S. Cuomo, V. Schiano D. Cola, F. Giampaolo, G. Rozza, M. Raissi, et al., Scientific machine learning through physics-informed neural networks: where we are and what’s next, J. Sci. Comput., 92 (2022), 88. https://doi.org/10.1007/s10915-022-01939-z

[15] J. Su, G. Tao, R. Li, S. Zhang, Physics-driven deep learning methods and numerically intractable “bad” Jaulent-Miodek equation, Chaos, 35 (2025), 063101. https://doi.org/10.1063/5.0264041

[16] P. Lan, J. Su, S. Zhang, Surrogate modeling for unsaturated infiltration via the physics and equality-constrained artificial neural networks, J. Rock Mech. Geotech. Eng., 16 (2024), 2282-2295. https://doi.org/10.1016/j.jrmge.2023.09.014

[17] S. Zhang, P. Lan, J. Su, Wave-packet behaviors of the defocusing nonlinear Schrodinger equation based on the modified physics-informed neural networks, Chaos, 31 (2021), 113107. https://doi.org/10.1063/5.0067260

[18] J. Su, S. Zhang, P. Lan, X. Chen, Explorations of certain nonlinear waves of the Boussinesq and Camassa-Holm equations using physics-informed neural networks, Proc. R. Soc. A, 480 (2024), 20230580. https://doi.org/10.1098/rspa.2023.0580

[19] S. Zhang, J. Deng, X. Li, Z. Zhao, J. Wu, W. Li, et al., Solving the one dimensional vertical suspended sediment mixing equation with arbitrary eddy diffusivity profiles using temporal normalized physics-informed neural networks, Phys. Fluids, 36 (2024), 017132. https://doi.org/10.1063/5.0179223

[20] X. Li, J. Wu, X. Tai, J. Xu, Y. Wang, Solving a class of multi-scale elliptic PDEs by Fourier-based mixed physics informed neural networks, J. Comput. Phys., 508 (2024), 113012. https://doi.org/10.1016/j.jcp.2024.113012

[21] X. Li, J. Wu, Y. Huang, Z. Ding, X. Tai, L. Liu, et al., Fourier-feature induced physics informed randomized neural network method to solve the biharmonic equation, J. Comput. Appl. Math., 468 (2025), 116635. https://doi.org/10.1016/j.cam.2025.116635

[22] S. Basir, I. Senocak, Physics and equality constrained artificial neural networks: application to forward and inverse problems with multi-fidelity data fusion, J. Comput. Phys., 463 (2022), 111301. https://doi.org/10.1016/j.jcp.2022.111301

[23] M. J. D. Powell, A method for nonlinear constraints in minimization problems, Optimization, 1969.

[24] A. Quarteroni, A. Valli, Domain Decomposition Methods for Partial Differential Equations , Oxford University Press, New York, 1999. https://doi.org/10.1093/oso/9780198501787.001.0001

[25] A. Toselli, O. Widlund, Domain Decomposition Methods - Algorithms and Theory , Springer Verlag, Berlin-Heidelberg-New York, 2005.

[26] A. D. Jagtap, E. Kharazmi, G. E. Karniadakis, Conservative physics-informed neural networks on discrete domains for conservation laws: applications to forward and inverse problems, Comput. Methods Appl. Mech. Eng., 365 (2020), 113028. https://doi.org/10.1016/j.cma.2020.113028

[27] A. D. Jagtap, G. E. Karniadakis, Extended physics-informed neural networks (XPINNs): a generalized space-time domain decomposition based deep learning framework for nonlinear partial differential equations, Commun. Comput. Phys., 28 (2020), 2002-2041. https://doi.org/10.4208/cicp.OA-2020-0164

[28] K. Shukla, A. D. Jagtap, G. Karniadakis, Parallel physics-informed neural networks via domain decomposition, J. Comput. Phys., 447 (2021), 110683. https://doi.org/10.1016/j.jcp.2021.110683

[29] W. Li, X. Xiang, Y. Xu, Deep domain decomposition method: Elliptic problems, Proc. Mach. Learn. Res., 107 (2020), 269-286. Available from: https://proceedings.mlr.press/v107/li20a.html.

[30] W. Li, Z. Wang, T. Cui, Y. Xu, X. Xiang, Deep domain decomposition methods: Helmholtz equation, Adv. Appl. Math. Mech., 15 (2023), 118-138. https://doi.org/10.4208/aamm.OA-2021-0305

[31] H. J. Yang, H. H. Kim, Iterative algorithms for partitioned neural network approximation to partial differential equations, Comput. Math. Appl., 170 (2024), 237-259. https://doi.org/10.1016/j.camwa.2024.07.007

[32] D. Jang, K. Kim, H. H. Kim, Partitioned neural network approximation for partial differential equations enhanced with Lagrange multipliers and localized loss functions, Comput. Methods Appl. Mech. Eng., 429 (2024), 117168. https://doi.org/10.1016/j.cma.2024.117168

[33] V. Dolean, A. Heinlein, S. Mishra, B. Moseley, Multilevel domain decomposition-based architectures for physics-informed neural networks, Comput. Methods Appl. Mech. Eng., 429 (2024), 117116. https://doi.org/10.1016/j.cma.2024.117116

[34] Q. Hu, S. Basir, I. Senocak, Non-overlapping Schwarz-type domain decomposition method for physics and equality constrained artificial neural networks, Comput. Methods Appl. Mech. Eng., 436 (2025), 117706. https://doi.org/10.1016/j.cma.2024.117706

[35] E. Olson, S. Won, X. Tu, Scalable two-level domain decomposition methods for the physics and equality constrained neural networks, Comput. Methods Appl. Mech. Eng., In press.

[36] M. J. Gander, H. Zhang, A class of iterative solvers for the Helmholtz equation: factorizations, sweeping preconditioners, source transfer, single layer potentials, polarized traces, and optimized Schwarz methods, SIAM Rev., 61 (2019), 3-76. https://doi.org/10.1137/16M109781X

[37] S. Basir, I. Senocak, An adaptive augmented Lagrangian method for training physics and equality constrained artificial neural networks, preprint, arXiv:2306.04904.

[38] V. Mercier, S. Gratton, P. Boudier, A coarse space acceleration of deep-DDM, preprint, arXiv:2112.03732.