Deep learning methods encounter significantly low-efficiency and low-accuracy challenges in effectively computing multi-scale multi-physics problems. In this study, an innovative higher-order multi-scale physics-informed randomized neural network (HOMS-PIRNN) framework is presented to efficiently and accurately simulate the dynamic thermo-mechanical coupling problems of composite materials, which integrates the benefits of HOMS-PIRNN. In the new framework, the higher-order multi-scale method decomposes the multi-scale multi-physics physical constraints for deep learning simulation into lower-order, higher-order microscopic physical constraints (microscopic cell equations) and macroscopic physical constraints (macroscopic homogenized equations), in which higher-order microscopic cell functions are capable of precisely depicting the phenomenon of intense oscillations at the microscale. Next, PIRNN are devised to mesh-free, efficiently and accurately solve microscopic cell functions and macroscopic homogenized solutions. Furthermore, the automatic differentiation technique of neural network is employed to construct high-accuracy multi-scale asymptotic solutions. Moreover, under appropriate assumptions, the error estimation of the HOMS-PIRNN method is obtained. Finally, various numerical experiments including two-dimensional and three-dimensional composite materials, as well as porous materials are carried out to verify the proposed method, not only showing it can outperform FEM and multi-scale method in terms of efficiency, but also illustrating its validity, especially for simulating large-scale material and long-time problems in terms of computational costs.
ChungETEfendievYLeungWT. An adaptive generalized multiscale discontinuous Galerkin method for high-contrast flow problems. Multiscale Model Simul2018; 16(3): 1227–1257.
9.
FuSChungEMaiT. Constraint energy minimizing generalized multiscale finite element method for nonlinear poroelasticity and elasticity. J Comput Phys2020; 417: 109569.
10.
HenningPMålqvistA. Localized orthogonal decomposition techniques for boundary value problems. SIAM J Sci Comput2014; 36(4): A1609–A1634.
11.
LiZHMaQCuiJ. Second-order two-scale finite element algorithm for dynamic thermo–mechanical coupling problem in symmetric structure. J Comput Phys2016; 314: 712–748.
12.
DongHCuiJNieY, et al. Second-order two-scale computational method for nonlinear dynamic thermo-mechanical problems of composites with cylindrical periodicity. Commun Comput Phys2017; 21(4): 1173–1206.
13.
DongHCuiJNieY, et al. Multiscale computational method for transient heat conduction problems of periodic porous materials with diverse periodic configurations in different subdomains. Appl Numer Math2019; 136: 215–234.
14.
WangXCaoLWongY. Multiscale computation and convergence for coupled thermoelastic system in composite materials. Multiscale Model Simul2015; 13(2): 661–690.
15.
YouHXiangMJingY, et al. A strain gradient phase field model for heterogeneous materials based on two-scale asymptotic homogenization. J Mech Phys Solids2025; 200: 106104.
16.
DongHZhengXCuiJ, et al. Multi-scale computational method for dynamic thermo-mechanical performance of heterogeneous shell structures with orthogonal periodic configurations. Comput Methods Appl Mech Eng2019; 354: 143–180.
17.
DongHCuiJNieY, et al. Multi-scale computational method for nonlinear dynamic thermo-mechanical problems of composite materials with temperature-dependent properties. Commun Nonlinear Sci Numer Simul2023; 118: 107000.
18.
YangZCuiJSunY, et al. Multiscale analysis method for thermo-mechanical performance of periodic porous materials with interior surface radiation. Int J Numer Methods Eng2016; 105(5): 323–350.
19.
DongHZhengXCuiJ, et al. Multiscale computational method for dynamic thermo-mechanical problems of composite structures with diverse periodic configurations in different subdomains. J Sci Comput2019; 79: 1630–1666.
20.
RaissiMPerdikarisPKarniadakisGE. Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J Comput Phys2019; 378: 686–707.
21.
EWYuB. The deep Ritz method: a deep learning-based numerical algorithm for solving variational problems. Commun Math Stat2018; 6(1): 1–12.
22.
SirignanoJSpiliopoulosK. DGM: a deep learning algorithm for solving partial differential equations. J Comput Phys2018; 375: 1339–1364.
23.
LiaoYMingP. Deep Nitsche method: deep Ritz method with essential boundary conditions. Commun Comput Phys2021; 29(5): 1365–1384.
24.
DongSLiZ. Local extreme learning machines and domain decomposition for solving linear and nonlinear partial differential equations. Comput Methods Appl Mech Eng2021; 387: 114129.
25.
DongSWangY. A method for computing inverse parametric PDE problems with random-weight neural networks. J Comput Phys2023; 489: 112263.
26.
DongSLiZ. A modified batch intrinsic plasticity method for pre-training the random coefficients of extreme learning machines. J Comput Phys2021; 445: 110585.
27.
LiYWangF. Local randomized neural networks methods for interface problems. arXiv [preprint]2023. DOI: 10.48550/arXiv.2308.03087.
HanJLeeY. A neural network approach for homogenization of multiscale problems. Multiscale Model Simul2023; 21(2): 716–734.
36.
JiangJWuJChenQ, et al. Physically informed deep homogenization neural network for unidirectional multiphase/multi-inclusion thermoconductive composites. Comput Methods Appl Mech Eng2023; 409: 115972.
37.
WuJJiangJChenQ, et al. Deep homogenization networks for elastic heterogeneous materials with two-and three-dimensional periodicity. Int J Solids Struct2023; 284: 112521.
38.
LinghuJDongHNieY, et al. Higher-order multi-scale deep Ritz method (HOMS-DRM) and its convergence analysis for solving thermal transfer problems of composite materials. Comput Mech2024; 75: 71–95.
39.
SantosoRDegenDCacaceM, et al. State-of-the-art physics-based machine learning for hydro-mechanical simulation in geothermal applications. In: European geothermal congress, Berlin, 17–2 October, 2022, pp. 1–10. Munich: Fraunhofer-Publica.
40.
CaoL. Multiscale asymptotic expansion and finite element methods for the mixed boundary value problems of second order elliptic equation in perforated domains. Numer Math2006; 103(1): 11–45.
41.
SukumarNSrivastavaA. Exact imposition of boundary conditions with distance functions in physics-informed deep neural networks. Comput Methods Appl Mech Eng2022; 389: 114333.
42.
GononL. Random feature neural networks learn black-scholes type PDEs without curse of dimensionality. J Mach Learn Res2023; 24(189): 1–51.
43.
De RyckTMishraSShangY, et al. Approximation theory and applications of randomized neural networks for solving high-dimensional PDEs. arXiv [preprint]2025. DOI: 10.48550/arXiv.2501.12145.
44.
DongHCuiJNieY, et al. Second-order two-scale computational method for damped dynamic thermo-mechanical problems of quasi-periodic composite materials. J Comput Appl Math2018; 343: 575–601.
45.
LinghuJGaoWDongH, et al. Higher-order multi-scale physics-informed neural network (HOMS-PINN) method and its convergence analysis for solving elastic problems of authentic composite materials. J Comput Appl Math2025; 456: 116223.
46.
ZuoHYangZCuiJ, et al. High-order models for hydro-mechanical coupling problems in multiscale porous media. Int J Numer Methods Eng2024; 125(11): e7456.
