Finite element neural network method for simulating two-dimensional partial differential equations and identifying parameters

Machine Learning


  • Niroomandi, S., Alfaro, I., Gonzalez, D., Cueto, E., Chinasta, F. Real-time simulation of surgery using low-order modeling and X-FEM technology. internal. J. Number. Method Biomed. engineering 28574–588 (2012).

    Google Scholar

  • Kuiper, W., Milde, A., Volkwein, S. Reduced Order Modeling (ROM) for Simulation and Optimization: Powerful Algorithms as a Key Enabler for Scientific Computing (Springer, 2018).

  • Barnett, D.S. Finite element analysis: from concept to application (Addison-Wesley Pub. Co., 1987).

  • Reddy, J. Overview of finite element method (McGraw-Hill, 1993).

  • Raissi, M., Perdicaris, P. & Karniadakis, GE Physics-based deep learning (part i): data-driven solutions of nonlinear partial differential equations. arXiv preprint arXiv:1711.10561 (2017).

  • Raissi, M., Perdicaris, P. & Karniadakis, GE Physics-based neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J. Compute. Physics. 378686–707 (2019).

    Google Scholar

  • Jenkiewicz, OC & Taylor, RL. Finite element method: solid mechanicsvol. 2 (Butterworth-Heinemann, 2000).

  • Meethal, RE et al. Neural networks enhanced with finite element methods for forward and inverse problems. Advanced model. Simul. engineering science. 106 (2023).

    Google Scholar

  • Dissanayake, MG & Phan-Thien, N. Neural network-based approximations for solving partial differential equations. common. numbers. method engineering 10195–201 (1994).

    Google Scholar

  • Kharazmi, E., Zhang, Z. & Karniadakis, GE Variational physics-based neural networks for solving partial differential equations. arXiv preprint arXiv:1912.00873 (2019).

  • Daubechies, I., DeVore, R., Foucart, S., Hanin, B. & Petrova, G. Nonlinear approximations and (deep) Relu networks. Structure approx. 55127–172 (2022).

    Google Scholar

  • Mhaskar, HN & Poggio, T. Function approximation with deep networks. arXiv preprint arXiv:1905.12882 (2019).

  • Jagtap, A.D., Kharazmi, E. & Karniadakis, G.E. Conservative physics-based neural networks over discrete domains of conservation laws: Applications to forward and inverse problems. Calculate. Applying the method. Mecha. engineering 365113028 (2020).

    Google Scholar

  • Kharazmi, E., Zhang, Z. & Karniadakis, GE hp-VPINN: Variational physics information neural network with domain decomposition. Calculate. Applying the method. Mecha. engineering 374113547 (2021).

    Google Scholar

  • Sunil, P. & Sills, RB FE-PINNS: Finite element-based physically informed neural networks for surrogate modeling. arXiv preprint arXiv:2412.07126 (2024).

  • Raissi, M. & Karniadakis, G. E. Hidden physical models: Machine learning for nonlinear partial differential equations. J. Compute. Physics. 357125–141. https://doi.org/10.1016/j.jcp.2017.11.039 (2017).

    Google Scholar

  • Raissi, M., Perdicaris, P. & Karniadakis, GE Physics-based neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J. Compute. Physics. 378686–707. https://doi.org/10.1016/j.jcp.2018.10.045 (2019).

    Google Scholar

  • McClenny, LD & Braga-Neto, UM Self-adaptive physics-informed neural networks. J. Compute. Physics. 474111722 (2023).

    Google Scholar

  • Abueidda, DW, Lu, Q. & Koric, S. A deep learning method for 3D solid mechanics based on meshless physics. internal. J. Number. Female. engineering 1227182–7201 (2021).

    Google Scholar

  • Wang, L., Liu, G., Wang, G. & Zhang, K. M-PINN: A mesh-based physics-informed neural network for linear elasticity problems in solid mechanics. internal. J. Number. Female. engineering 125e7444 (2024).

    Google Scholar

  • Wang, S., Teng, Y. & Perdicaris, P. Understanding and mitigating gradient pathology in physics-based neural networks. arXiv preprint arXiv:2001.04536 (2020).

  • Fuks, O. & Tchelepi, H.A. Physics limitations influenced machine learning of nonlinear two-phase transport in porous media. J. Mach. learn. model. Calculate.1 (2020).

  • Liu, C. & Wu, H. cv-PINN: Efficient learning of variational physics-based neural networks with domain decomposition. Extreme mecha. Let. 63102051 (2023).

    Google Scholar

  • Anandh, T., Ghose, D., Jain, H., Ganesan, S. FastVPINN: Tensor-driven acceleration of VPINN for complex geometries. arXiv preprint arXiv:2404.12063 (2024).

  • Anand, T. et al. An efficient hp variational PINN framework for incompressible Navier-Stokes equations. arXiv preprint arXiv:2409.04143 (2024).

  • Abda, M., Hamedi, M., Piollet, E., Blake, C. & Gosselin, FP Finite element neural network methods: A one-dimensional study. next resolution 2100885. https://doi.org/10.1016/j.nexres.2025.100885 (2025).

    Google Scholar

  • Yu, B. et al. Deep Ritz Method: A deep learning-based numerical algorithm for solving variational problems. common. Mathematics. statistics 61–12 (2018).

    Google Scholar

  • Samaniego, E. et al. An energetic approach to solving partial differential equations in computational mechanics with machine learning: Concepts, implementation, and applications. Calculate. Applying the method. Mecha. engineering 362112790 (2020).

    Google Scholar

  • Nguyen-Thanh, VM, Zhuang, X. & Rabczuk, T. Deep energy method for finite deformation hyperelasticity. EUR. J. Mech.-A/Solid 80103874 (2020).

    Google Scholar

  • Fuhg, JN & Bouklas, N. A mixed deep energy method for resolving lumped features in finite strain hyperelasticity. J. Compute. Physics. 451110839 (2022).

    Google Scholar

  • Dong, Y., Liu, T., Li, Z. & Qiao, P. DeepFEM: a new element-based deep learning approach for solving nonlinear partial differential equations in computational solid mechanics. J.Eng.Mecha. 14904022102 (2023).

    Google Scholar

  • Wang, X., Yin, Z.-Y., Wu, W., Zhu, H.-H. Neural network augmented differentiable finite element method for boundary value problems. internal. J. Mech. Science. 285109783 (2025).

    Google Scholar

  • Cai, S., Wang, Z., Wang, S., Perdicaris, P. & Karniadakis, GE Physics-based neural networks for heat transfer problems. J. Heat transfer 143060801 (2021).

    Google Scholar

  • Zhang, N., Xu, K., Yin, Z.Y., and Li, K.-Q. & Jin, Y.-F. A finite element integrated neural network framework for elastic and elastoplastic solids. Calculate. Applying the method. Mecha. engineering 433117474 (2025).

    Google Scholar

  • Abadi, M. et al. Tensorflow: Large-scale machine learning on heterogeneous distributed systems. arXiv preprint arXiv:1603.04467 (2016).

  • Xu, R., Zhang, D., Rong, M. & Wang, N. Weak-form theory-guided neural network (TgNN-wf) for deep learning of subsurface single- and two-phase flows. J. Compute. Physics. 436110318 (2021).

    Google Scholar

  • Rantangen, H.P. & Mardal, K.-A. Introduction to numerical methods for variational problemsvol. 21 (Springer Nature, 2019).

  • Bochev, MD & Gunzberger, MD least squares finite element methodvol. 166 (Springer Science & Business Media, 2009).

  • Burgers, JM A mathematical model to explain the theory of turbulence. Advanced applied mechanics. 1171–199 (1948).

    Google Scholar

  • Kingma, DP Adam: Stochastic optimization techniques. arXiv preprint arXiv:1412.6980 (2014).

  • Liu, D. & Wang, Y. A dual-dimer method for training physically constrained neural networks on minimax architectures. Neural network. 136112–125 (2021).

    Google Scholar

  • Liu, DC and Nocedal, J. On limited memory BFGS methods for large-scale optimization. Mathematics. program. 45503–528 (1989).

    Google Scholar

  • Basdevant, C. et al. Spectral solution and finite difference decomposition of the Berger equation. Calculate. fluid 1423–41 (1986).

    Google Scholar

  • Orlandi, P. Fluid flow phenomena: Numerical toolkitvol. 55 (Springer Science & Business Media, 2012).

  • Jagtap, A.D., Kawaguchi, K. & Karniadakis, G.E. Adaptive activation functions accelerate convergence in physics-based deep neural networks. J. Compute. Physics. 404109136 (2020).

    Google Scholar

  • Xiao, Y. et al. A least-squares finite difference-based physics-based neural network for steady incompressible flows. Calculate. Mathematics. application 17533–48 (2024).

    Google Scholar

  • Marchi, CH, Suero, R. & Araki, LK Lid-driven square cavity flow: Numerical solution using a 1024 x 1024 grid. J. Blaz Social Mecha. Science. engineering 31186–198 (2009).

    Google Scholar

  • Ghia, U., Ghia, KN & Shin, C. High-resolution solutions for incompressible flows using Navier-Stokes equations and multigrid methods. J. Compute. Physics. 48387–411 (1982).

    Google Scholar

  • Botella, O. and Peyret, R. Benchmark spectral results for lid-driven cavity flow. Calculate. fluid 27421–433 (1998).

    Google Scholar

  • Alphonius, A. et al. Lethe 1.0: Open source, high-performance, high-order computational fluid dynamics software for single-phase and multiphase flows. Available at SSRN 5090483 (2025).

  • Chandan, K. et al. Physics-based Hermitian neural network for wet porous fins under local thermal nonequilibrium conditions: Application of clique polynomial method. EUR. Physics. J.Specification top. 1-21 (2024).

  • Chandan, K. et al. Radiative heat transfer analysis of concave porous fins under local thermal nonequilibrium conditions: Application of clique polynomial method and physics-based neural network. applied mathematics. Mecha. 451613–1632 (2024).

    Google Scholar

  • Kumar, C. et al. Physics-based machine learning predictions for thermal analysis in convective-radiative concave fins with periodic boundary conditions. Zam-J. Applied Mathematics. Mech./Zeitschrift für Angewandte Mathematik und Mechanik 104e202300712 (2024).

    Google Scholar

  • Chandan, K. et al. Physics-based optimized neural network for analyzing the radiative-convective thermal performance of tilted wavy porous fins. case stud. thermal engineering 64105423 (2024).

    Google Scholar

  • Oommen, V. & Srinivasan, B. Solving inverse heat transfer problems without surrogate models: A fast, data-sparse, physics-based neural network approach. J. Compute. Information science. engineering twenty two041012 (2022).

    Google Scholar

  • MayaHTT. TMG heat flow solver. https://help.mayahtt.com/tmg/index.html (2026). Accessed in 2026.

  • Liu, L. et al. Discontinuity computing using physics-based neural networks. J.Sci.Calculate. 9822 (2024).

    Google Scholar

  • Jusène, C. & Remacre, J.-F. Gmsh: 3D finite element mesh generator with built-in pre- and post-processing capabilities. internal. J. Number. Female. engineering 791309–1331 (2009).

    Google Scholar



  • Source link