A collocation point sampling method for solving the Helmholtz equation using physics-informed neural networks

Authors

SHAN Lianpu, Key Laboratory of Deep Petroleum Intelligent Exploration and Development, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, China; College of Earth and Planetary Sciences, University of Chinese Academy of Sciences, Beijing 100049, ChinaFollow
YU Caixia, Key Laboratory of Deep Petroleum Intelligent Exploration and Development, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, China
WANG Yanfei, Key Laboratory of Deep Petroleum Intelligent Exploration and Development, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, China; College of Earth and Planetary Sciences, University of Chinese Academy of Sciences, Beijing 100049, ChinaFollow

Abstract

Background In the field of seismic wavefield simulation, physics-informed neural networks (PINNs) have emerged as a new method for efficiently solving the Helmholtz equation due to their characteristic of grid-free computation. However, for the forward modeling of complex medium models using the Helmholtz equation, traditional uniform grid-based methods for collocation point sampling are insufficient to efficiently provide gradient information, leading to prolonged training time and high resource consumption. Objective and Method To improve the forward modeling efficiency and accuracy based on PINNs and overcome the anomalous concentration of collocation points in the seismic source during dynamic sampling, this study proposed a hybrid residual-based adaptive distribution (RAD-H) method and then compared this method with six fixed and four dynamic sampling methods. In the RAD-H method, fixed sampling is employed for the seismic source area to ensure the capture of the source features, and residual-based adaptive distribution (RAD) is used for other areas to ensure the consistency between the resampled collocation points and the spatial sampling of the high-residual target area. Results Numerical experiments indicate that for a homogeneous medium model, the RAD-H method reduced the L2 error of the wavefield by 3.050% using only 18.57% of collocation points compared to the classical baseline method. Moreover, the RAD-H method exhibited a model training time of approximately 4192.5 s, suggesting an improvement of 231.74% in computational efficiency compared to the classical baseline method. For the Marmousi2 model, the RAD-H strategy remained the error within the same order of magnitude as the classical baseline method while reducing the number of collocation points by 57.75%. Conclusions The RAD-H method overcomes oversampling for seismic source areas and undersampling for other areas of other dynamic sampling methods, significantly enhancing the sampling efficiency of collocation points of PINNs during the forward modeling involving seismic sources. This study holds significant implications for efficient, high-precision wavefield simulation of complex geological models.

Keywords

physics-informed neural networks (PINNs), Helmholtz equation, forward modeling, collocation point, adaptive distribution

DOI

10.12363/issn.1001-1986.25.03.0197

Recommended Citation

SHAN Lianpu, YU Caixia, WANG Yanfei, et al. (2025) "A collocation point sampling method for solving the Helmholtz equation using physics-informed neural networks," Coal Geology & Exploration: Vol. 53: Iss. 9, Article 17.
DOI: 10.12363/issn.1001-1986.25.03.0197
Available at: https://cge.researchcommons.org/journal/vol53/iss9/17

Reference

[1] 方修政，姚刚，钮凤林，等. 波场模拟有限差分参数优化选取[J]. 地球物理学报，2023，66(6)：2520−2533.

FANG Xiuzheng，YAO Gang，NIU Fenglin，et al. Estimating optimal parameters of finite–difference scheme for wavefield modeling[J]. Chinese Journal of Geophysics，2023，66(6)：2520−2533.

[2] 苏波，李怀良，刘少林，等. 修正辛格式有限元法的地震波场模拟[J]. 地球物理学报，2019，62(4)：1440−1452.

SU Bo，LI Huailiang，LIU Shaolin，et al. Modified symplectic scheme with finite element method for seismic wavefield modeling[J]. Chinese Journal of Geophysics，2019，62(4)：1440−1452.

[3] 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.

[4] 李野，陈松灿. 基于物理信息的神经网络：最新进展与展望[J]. 计算机科学，2022，49(4)：254−262.

LI Ye，CHEN Songcan. Physics–informed neural networks：Recent advances and prospects[J]. Computer Science，2022，49(4)：254−262.

[5] 李航，孙宇航，李佳慧，等. 基于波动方程的地震波数值模拟研究综述[J]. 吉林大学学报(地球科学版)，2025，55(2)：627−645.

LI Hang，SUN Yuhang，LI Jiahui，et al. A comprehensive review of numerical simulation of seismic waves based on wave equation[J]. Journal of Jilin University (Earth Science Edition)，2025，55(2)：627−645.

[6] BIN WAHEED U，HAGHIGHAT E，ALKHALIFAH T，et al. PINNeik：Eikonal solution using physics–informed neural networks[J]. Computers & Geosciences，2021，155：104833.

[7] 潘登，唐杰，范忠豪，等. 基于自适应损失均衡梯度增强的物理信息神经网络微地震定位[J]. 石油地球物理勘探，2025，60(3)：630−642.

PAN Deng，TANG Jie，FAN Zhonghao，et al. Adaptive–loss–weighting gradient–enhanced physics–informed neural network for microseismic localization[J]. Oil Geophysical Prospecting，2025，60(3)：630−642.

[8] 都国宁，谭军，宋鹏，等. 基于物理信息驱动神经网络的三维初至波旅行时计算方法[J]. 石油地球物理勘探，2023，58(1)：9−20.

DU Guoning，TAN Jun，SONG Peng，et al. 3D traveltime calculation of first arrival wave using physics–informed neural network[J]. Oil Geophysical Prospecting，2023，58(1)：9−20.

[9] KARIMPOULI S，TAHMASEBI P. Physics informed machine learning：Seismic wave equation[J]. Geoscience Frontiers，2020，11(6)：1993−2001.

[10] BIN WAHEED U. Kronecker neural networks overcome spectral bias for PINN–based wavefield computation[J]. IEEE Geoscience and Remote Sensing Letters，2022，19：8029805.

[11] SONG Chao，ALKHALIFAH T，BIN WAHEED U. A versatile framework to solve the Helmholtz equation using physics–informed neural networks[J]. Geophysical Journal International，2022，228(3)：1750−1762.

[12] SONG Chao，WANG Yanghua. Simulating seismic multifrequency wavefields with the Fourier feature physics–informed neural network[J]. Geophysical Journal International，2023，232(3)：1503−1514.

[13] XU Yiran，LI Jingye，CHEN Xiaohong. Physics informed neural networks for velocity inversion[C]//SEG International Exposition and Annual Meeting. San Antonio：Society of Exploration Geophysicists，2019：SEG–2019–3216823.

[14] RASHT–BEHESHT M，HUBER C，SHUKLA K，et al. Physics–informed neural networks (PINNs) for wave propagation and full waveform inversions[J]. Journal of Geophysical Research：Solid Earth，2022，127(5)：e2021JB023120.

[15] 邹婧博，刘财，赵鹏飞. 地震波正演问题的物理信息神经网络算法研究进展[J]. 地球物理学进展，2023，38(1)：430−448.

ZOU Jingbo，LIU Cai，ZHAO Pengfei. Research progress of physics–informed neural network in seismic wave modeling[J]. Progress in Geophysics，2023，38(1)：430−448.

[16] WU Yuanqing，SUN Shuyu. Enhancing the accuracy of physics–informed neural network surrogates in flash calculations using sparse grid guidance[J]. Fluid Phase Equilibria，2024，578：113984.

[17] JO C H，SHIN C，SUH J H. An optimal 9–point，finite–difference，frequency–space，2–D scalar wave extrapolator[J]. Geophysics，1996，61(2)：529−537.

[18] SOBOL I M. On the distribution of points in a cube and the approximate evaluation of integrals[J]. USSR Computational Mathematics and Mathematical Physics，1967，7(4)：86−112.

[19] HALTON J H. On the efficiency of certain quasi–random sequences of points in evaluating multi–dimensional integrals[J]. Numerische Mathematik，1960，2(1)：84−90.

[20] HAMMERSLEY J M，HANDSCOMB D C. Monte carlo methods[M]. London：Methuen & Wiley，1964.

[21] WANG G，FANG Q，WANG J，et al. Estimation of load for tunnel lining in elastic soil using physics–informed neural network[J]. Computer–Aided Civil and Infrastructure Engineering，2024，39(17)：2701−2718.

[22] IWASAKI Y，LAI C Y. One–dimensional ice shelf hardness inversion：Clustering behavior and collocation resampling in physics–informed neural networks[J]. Journal of Computational Physics，2023，492：112435.

[23] NING Luyuan，CAI Zhenwei，DONG Han，et al. A peridynamic–informed neural network for continuum elastic displacement characterization[J]. Computer Methods in Applied Mechanics and Engineering，2023，407：115909.

[24] REN Xiaodan，LYU Xianrui. Mixed form based physics–informed neural networks for performance evaluation of two–phase random materials[J]. Engineering Applications of Artificial Intelligence，2024，127：107250.

[25] ZHANG Xiaoping，ZHU Yan，WANG Jing，et al. GW–PINN：A deep learning algorithm for solving groundwater flow equations[J]. Advances in Water Resources，2022，165：104243.

[26] CHEN Lin，LI Ben，LUO Chenyi，et al. WaveNets：Physics–informed neural networks for full–field recovery of rotational flow beneath large–amplitude periodic water waves[J]. Engineering with Computers，2024，40(5)：2819−2839.

[27] ALKHADHR S，ALMEKKAWY M. Physics–informed neural networks with resampling technique to model ultrasound wave propagation of a multi–element transducer[C]//2022 IEEE International Ultrasonics Symposium (IUS). Venice：IEEE，2022：1–4.

[28] ZHOU Wen，MIWA S，OKAMOTO K. Advancing fluid dynamics simulations：A comprehensive approach to optimizing physics–informed neural networks[J]. Physics of Fluids，2024，36(1)：013615.

[29] LIU Haiyi，ZHANG Yabin，WANG Lei. Pre–training physics–informed neural network with mixed sampling and its application in high–dimensional systems[J]. Journal of Systems Science and Complexity，2024，37(2)：494−510.

[30] DAI Qianwei，ZHOU Wei，HE Run，et al. A data assimilation methodology to analyze the unsaturated seepage of an earth–rockfill dam using physics–informed neural networks based on hybrid constraints[J]. Water，2024，16(7)：1041.

[31] WANG Tian，BAI Zhaowei，ZHAO Haixia. Effects of sampling strategies and optimization methods on solving wave equation with physics–informed neural network[C]//IGARSS 2023–2023 IEEE International Geoscience and Remote Sensing Symposium. Pasadena：IEEE，2023：2215–2218.

[32] HANNA J M，AGUADO J V，COMAS–CARDONA S，et al. Residual–based adaptivity for two–phase flow simulation in porous media using physics–informed neural networks[J]. Computer Methods in Applied Mechanics and Engineering，2022，396：115100.

[33] WU Chenxi，ZHU Min，TAN Qinyang，et al. A comprehensive study of non–adaptive and residual–based adaptive sampling for physics–informed neural networks[J]. Computer Methods in Applied Mechanics and Engineering，2023，403：115671.

[34] GAO Zhiwei，YAN Liang，ZHOU Tao. Failure–informed adaptive sampling for PINNs[J]. SIAM Journal on Scientific Computing，2023，45(4)：A1971−A1994.

[35] ESHKOFTI K，HOSSEINI S M. A new modified deep learning technique based on physics–informed neural networks (PINNs) for the shock–induced coupled thermoelasticity analysis in a porous material[J]. Journal of Thermal Stresses，2024，47(6)：798−825.

[36] HOU Jie，LI Ying，YING Shihui. Enhancing PINNs for solving PDEs via adaptive collocation point movement and adaptive loss weighting[J]. Nonlinear Dynamics，2023，111(16)：15233−15261.

[37] LIU Jianan，HOU Qingzhi，WEI Jianguo，et al. ESR–PINNs：Physics–informed neural networks with expansion–shrinkage resampling selection strategies[J]. Chinese Physics B，2023，32(7)：070702.

[38] ZONG Yifei，HE Qizhi，TARTAKOVSKY A M. Improved training of physics–informed neural networks for parabolic differential equations with sharply perturbed initial conditions[J]. Computer Methods in Applied Mechanics and Engineering，2023，414：116125.

[39] MAO Zhiping，MENG Xuhui. Physics–informed neural networks with residual/gradient–based adaptive sampling methods for solving partial differential equations with sharp solutions[J]. Applied Mathematics and Mechanics (English Edition)，2023，44(7)：1069−1084.

[40] LI Jinwu，LONG Xiangyun，DENG Xinyang，et al. A principled distance–aware uncertainty quantification approach for enhancing the reliability of physics–informed neural network[J]. Reliability Engineering & System Safety，2024，245：109963.

[41] AIKAWA Y，UEDA N，TANAKA T. Improving the efficiency of training physics–informed neural networks using active learning[J]. New Generation Computing，2024，42(4)：739−760.

[42] JIANG Xinchao，WANG Xin，WEN Ziming，et al. Practical uncertainty quantification for space–dependent inverse heat conduction problem via ensemble physics–informed neural networks[J]. International Communications in Heat and Mass Transfer，2023，147：106940.