Deep neural operators have emerged as a promising paradigm for efficiently solving parametric partial differential equations (PDEs) with random parameters via data-driven surrogate modeling.However, directly training global neural operators for large-scale, high-dimensional problems usually suffers from prohibitive computational costs and limited generalization capabilities.To address these challenges, this paper proposes a novel computational framework that couples nonoverlapping domain decomposition methods (DDM) with operator learning.This approach partitions the global computational domain into several local subdomains and utilizes local Karhunen-Loève (KL) expansions to effectively reduce the parameter dimensionality.In the training phase, each local neural operator is trained independently to learn the mapping from local parameters and interface conditions to the local solution.In the prediction phase, an optimization algorithm based on interface constraints is further proposed, enabling all local operators to infer the global approximation rapidly and in parallel.Compared with the global neural operator, domain decomposition significantly reduces the fitting difficulty of surrogate models and achieves higher prediction accuracy at lower computational cost.A numerical example on a two-dimensional stochastic diffusion equation demonstrates that the proposed local operator requires only about 20% of the network parameters to achieve a better and more robust approximation performance, with the relative error reduced by approximately 34%.Furthermore, its localized structure naturally supports parallel computing, providing a new paradigm for the efficient solution of large-scale parameterized PDEs.
XiuD, KarniadakisG E.Modeling uncertainty in flow simulations via generalized polynomial chaos [J].J.ComputPhys, 2003, 187(1): 137-167.
[5]
XingW W, TriantafyllidisV, ShahA A, et al.Manifold learning for the emulation of spatial fields from computational models [J].JComputPhys, 2016, 326: 666-690.
[6]
RaissiM, PerdikarisP, KarniadakisG E.Numerical Gaussian processes for time-dependent and nonlinear partial differential equations [J].SIAM J Sci Comput, 2018, 40(1): A172-A198.
ZhuY, ZabarasN.Bayesian deep convolutional encoder–decoder networks for surrogate modeling and uncertainty quantification [J].J ComputPhys, 2018, 366: 415-447.
[9]
LuL, JinP, PangG, et al.Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators [J].Nat Mach Intell, 2021, 3(3): 218-229.
[10]
LiZ, KovachkiN, AzizzadenesheliK, et al.Fourier neural operator for parametric partial differential equations [PP/OL].(2021-05-17)[2025-10-09]
[11]
LiZ, HuangD Z, LiuB, et al.Fourier neural operator with learned deformations for PDEs on general geometries [J].J Mach Learn Res, 2023, 24(388): 1-26.
[12]
WangS, WangH, PerdikarisP.Learning the solution operator of parametric partial differential equations with physics-informed DeepONets [J].Sci Adv, 2021, 7(40): eabi8605.
[13]
LiZ, ZhengH, KovachkiN, et al.Physics-informed neural operator for learning partial differential equations [J].ACM/IMS J Data Sci, 2024, 1(3): 1-27.
[14]
ToselliA, WidlundO.Domain decomposition methods-algorithms and theory [M].Berlin:Springer, 2004.
[15]
XuZ, LiaoQ, LiJ.Domain-decomposed Bayesian inversion based on local Karhunen-Loève expansions [J].J Comput Phys, 2024, 504: 112856.
ChenY, JakemanJ, GittelsonC, et al.Local polynomial chaos expansion for linear differential equations with high dimensional random inputs [J].SIAM J Sci Comput, 2015, 37(1): A79-A102.
[23]
CaflischR E.Monte Carlo and quasi-Monte Carlo methods [J].Acta Numer, 1998, 7: 1-49.
JolliffeI T, CadimaJ.Principal component analysis: a review and recent developments [J].Philos Trans R Soc A Math Phys Eng Sci, 2016, 374(2065): 20150202.
[26]
ElmanH C, RamageA, SilvesterD J.IFISS: A computational laboratory for investigating incompressible flow problems [J].SIAM Rev, 2014, 56(2): 261-273.
[27]
ElmanH C, SilvesterD J, WathenA J.Finite elements and fast iterative solvers: With applications in incompressible fluid dynamics [M].Oxford: Oxford University Press, 2014.