文章详情

题名 · 作者 · 摘要 · 关键词 · DOI · 参考文献

HT-PeRCNN:基于Hessian矩阵迹权重的物理编码递归卷积神经网络

伏倩

(中国石油化工股份有限公司胜利油田分公司滨南采油厂 信息化服务中心,山东 滨州 256600)

栏目:人工智能 · 第 2025 年第 11 期 · 页码:25-32,37 · 2096-4706(2025)11-0025-09 · 中图分类号:TP391.4 · 文献标识码:A · DOI:10.19850/j.cnki.2096-4706.2025.11.006

摘要

近年来,物理信息神经网络作为一种基于深度学习的偏微分方程求解方法,在多个领域取得了显著进展。然而,物理信息神经网络仍然存在训练效率低以及推理速度慢等问题。针对这些问题,该研究提出了一种基于Hessian矩阵迹权重的物理编码递归卷积神经网络(Physics-encoded Recurrent Convolutional Neural Network, PeRCNN)改进方法,HT-PeRCNN。该方法利用Hessian矩阵迹作为权重调节因子,以优化损失函数的权重分配,提高了模型的稳定性和外推能力。实验结果表明,与PeRCNN相比,HT-PeRCNN在多个PDE求解任务中将方程解的精度提高了50%。

关键词

偏微分方程;物理信息神经网络;物理编码递归卷积神经网络;Hessian矩阵迹权重

参考文献

  1. [1] CAI S,MAO Z,WANG Z,et al. Physics-Informed Neural Networks (PINNs) for Fluid Mechanics: A Review [J].Acta Mechanica Sinica,2021,37(12):1727-1738.
  2. [2] CAI S,WANG Z,WANG S,et al. Physics-Informed Neural Networks for Heat Transfer Problems [J/OL].Journal of Heat Transfer,2021,143(6):060801(2021-04-21).https://doi.org/10.1115/1.4050542.
  3. [3] WOLFF T,CARRILLO H,MARTI L,et al. Assessing Physics-Informed Neural Networks in Ocean Modelling and Climate Change Applications [R/OL].(2021-05-07).https://iclr.cc/virtual/2021/workshop/2123.
  4. [4] ANDERSON J D,WENDT J. Computational Fluid Dynamics [M].Cham:Springer,1995.
  5. [5] FRIES T P,BELYTSCHKO T. The Extended/Generalized Finite Element Method: An Overview of the Method and Its Applications [J].International Journal for Numerical Methods in Engineering,2010,84(3):253-304.
  6. [6] SHU C W. High-Order Finite Difference and Finite Volume WENO Schemes and Discontinuous Galerkin Methods for CFD [J].International Journal of Computational Fluid Dynamics,2003,17(2):107-118.
  7. [7] WEINAN E,HAN J,JENTZEN A,et al. Deep Learning-Based Numerical Methods for High-Dimensional Parabolic Partial Differential Equations and Backward Stochastic Differential Equations [J].Communications in Mathematics and Statistics,2017,5(4):349-380.
  8. [8] RAISSI M,PERDIKARIS P,KARNIADAKIS G E. Physics-Informed Neural Networks: A Deep Learning Framework for Solving Forward and Inverse Problems Involving Nonlinear Partial Differential Equations [J].Journal of Computational Physics,2019,378:686-707.
  9. [9] 田十方,李彪.基于梯度优化物理信息神经网络求解复杂非线性问题 [J].物理学报,2023,72(10):15-25.
  10. [10] JAGTAP A D,KARNIADAKIS G E. Extended Physics-Informed Neural Networks (XPINNs): A Generalized Space-Time Domain Decomposition Based Deep Learning Framework for Nonlinear Partial Differential Equations [J].Communications in Computational Physics,2020,28(5):2002-2041.
  11. [11] JIN X,CAI S,LI H,et al. NSFnets (Navier-Stokes Flow Nets): Physics-Informed Neural Networks for the Incompressible Navier-Stokes Equations [J/OL].Journal of Computational Physics,2021,426:109951(2021-01-05).https://doi.org/10.1016/j.jcp.2020.109951.
  12. [12] SARABIAN M,BABAEE H,LAKSARI K. Physics-Informed Neural Networks for Brain Hemodynamic Predictions Using Medical Imaging [J].IEEE Transactions on Medical Imaging,2022,41(9):2285-2303.
  13. [13] HAN J,XUE L,WEI Y,et al. Physics-Informed Neural Network-Based Petroleum Reservoir Simulation with Sparse Data Using Domain Decomposition [J].Petroleum Science,2023,20(6):3450-3460.
  14. [14] 徐宝昌,张学智,王雅欣,等.用于两相流环空压力预测的自适应物理信息神经网络模型 [J].石油学报,2023,44(3):545-555.
  15. [15] DU J,ZHENG J,LIANG Y,et al. Deeppipe:A Two Stage Physics Informed Neural Network for Predicting Mixed Oil Concentration Distribution [J/OL].Energy,2023,276:127452(2023-04-24).https://doi.org/10.1016/j.energy.2023.127452.
  16. [16] RAO C,REN P,WANG Q,et al. Encoding Physics to Learn Reaction–Diffusion Processes [J].Nature Machine Intelligence,2023,5(7):765-779.
  17. [17] SHI X,CHEN Z,WANG H,et al. Convolutional LSTM Network:A Machine Learning Approach for Precipitation Nowcasting [C]//NIPS'15: Proceedings of the 29th International Conference on Neural Information Processing Systems-Volume 1.Cambridge:MIT Press,2015:802-810.
  18. [18] KINGMA D P,BA J. Adam: A Method for Stochastic Optimization [J/OL].arXiv:1412.6980 [cs.LG].(2024-12-22).https://arxiv.org/abs/1412.6980.
  19. [19] FENG X,PROHL A. Numerical Analysis of the Allen-Cahn Equation and Approximation for Mean Curvature Flows [J].Numerische Mathematik,2003,94:33-65.
  20. [20] MCCLENNY L,BRAGA-NETO U. Self-Adaptive Physics-Informed Neural Networks Using a Soft Attention Mechanism [J/OL].arXiv:2009.04544 [cs.LG].(2024-06-18).https://arxiv.org/abs/2009.04544.
  21. [21] BONKILE M P,AWASTHI A,LAKSHMI C,et al. A Systematic Literature Review of Burgers’ Equation with Recent Advances [J].Pramana,2018,90:1-21.
  22. [22] KRUPA M,SANDSTEDE B,SZMOLYAN P. Fast and Slow Waves in the FitzHugh-Nagumo Equation [J].Journal of Differential Equations,1997,133(1):49-97.
  23. [23] REN P,RAO C,LIU Y,et al. PhyCRNet: Physics Informed Convolutional Recurrent Network for Solving Spatiotemporal PDEs [J/OL].Computer Methods in Applied Mechanics and Engineering,2022,389:114399(2021-12-18).https://doi.org/10.1016/j.cma.2021.114399.

作者简介

伏倩(1998.04—),女,汉族,宁夏固原人,助理工程师,学士学位,研究方向:偏微分方程智能化求解方法、智能油田建设。