Simulation and parameter identification of groundwater flow model basedon PINNs algorithms

ZHANG Sheng; LAN Peng; SU Jingjing; XIONG Haibin

doi:10.11779/CJGE20211138

Volume 45 Issue 2

Turn off MathJax

Article Contents

Abstract

References

Supplements

Chinese Journal of Geotechnical Engineering > 2023 > 45(2): 376-383. > DOI: 10.11779/CJGE20211138

ZHANG Sheng, LAN Peng, SU Jingjing, XIONG Haibin. Simulation and parameter identification of groundwater flow model basedon PINNs algorithms[J]. Chinese Journal of Geotechnical Engineering, 2023, 45(2): 376-383. DOI: 10.11779/CJGE20211138

Citation:

PDF (3031 KB)

Simulation and parameter identification of groundwater flow model basedon PINNs algorithms

School of Civil Engineering, Central South University, Changsha 410075, China

More Information

Received Date: September 25, 2021
Available Online: February 23, 2023

Graphical Abstract

Abstract

Abstract

The simulation of the Darcy velocity (forward problem) and the identification of seepage parameters (backward problem) in the groundwater flow model are of significance to practical projects, while at present, few algorithms can be used to simultaneously tackle these two problems. The physics-informed neural networks (PINNs) algorithms with the hard constraints are introduced for investigating these two problems at the same time. For the forward problem, two methods are established for deriving the Darcy velocity. One is to address the groundwater head and Darcy velocity concurrently by coupling the seepage flow equation with the Darcy's law (PINNs-H-I), and the other is to calculate the groundwater head first and then solve the Darcy velocity by automatic differentiation (PINNs-H-II). For the backward problem, the PINNs algorithms for the single and multi-physical field neural network models are used to identify the seepage parameters of homogeneous and non-homogeneous seepages, respectively. Furthermore, several examples are presented, and the results show that the hard-constraint PINNs algorithms exhibit better performances for the forward and backward problems compared with the soft-constraint ones. In addition, it is noted that PINNs-H-II possesses higher calculation accuracy, and both the PINNs algorithms for the single and multi-physical field neural network models can accurately identify the seepage parameters in the homogeneous and non-homogeneous seepage.
- groundwater numerical simulation,
- seepage parameter identification,
- physics-informed neural network,
- hard constraint

FullText(HTML)

References (32)

References

[1]	谢一凡, 吴吉春, 王益, 等. 一种模拟节点达西流速的多尺度有限元-有限元模型[J]. 岩土工程学报, 2022, 44(1): 107-114, 202. https://www.cnki.com.cn/Article/CJFDTOTAL-YTGC202201010.htm XIE Yifan, WU Jichun, WANG Yi, et al. Multiscale finite element-finite element model for simulating nodal Darcy velocity[J]. Chinese Journal of Geotechnical Engineering, 2022, 44(1): 107-114, 202. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-YTGC202201010.htm
[2]	谢一凡, 吴吉春, 薛禹群, 等. 一种模拟节点达西渗透流速的三次样条多尺度有限单元法[J]. 岩土工程学报, 2015, 37(9): 1727-1732. https://www.cnki.com.cn/Article/CJFDTOTAL-YTGC201509030.htm XIE Yifan, WU Jichun, XUE Yuqun, et al. Cubic-spline multiscale finite element method for simulation of nodal Darcy velocities in aquifers[J]. Chinese Journal of Geotechnical Engineering, 2015, 37(9): 1727-1732. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-YTGC201509030.htm
[3]	李守巨, 上官子昌, 刘迎曦, 等. 地下水渗流模型参数识别的模拟退火算法[J]. 岩石力学与工程学报, 2005(S1): 5031-5036. https://cpfd.cnki.com.cn/Article/CPFDTOTAL-ZGYJ200508002002.htm LI Shouju, SHANGGUAN Zichang, LIU Yingxi, et al. Parameter identification procedure for groundwater flow model with simulated annealing[J]. Chinese Journal of Rock Mechanics and Engineering, 2005(S1): 5031-5036. (in Chinese) https://cpfd.cnki.com.cn/Article/CPFDTOTAL-ZGYJ200508002002.htm
[4]	周凌峰, 王媛, 冯迪. 求解非均质渗流场的改进数值流形方法[J]. 岩土工程学报, 2021, 43(7): 1288-1296, 1377. https://www.cnki.com.cn/Article/CJFDTOTAL-YTGC202107018.htm ZHOU Lingfeng, WANG Yuan, FENG Di. An improved numerical manifold method for solving heterogeneous seepage problem[J]. Chinese Journal of Geotechnical Engineering, 2021, 43(7): 1288-1296, 1377. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-YTGC202107018.htm
[5]	YEH G T. On the computation of Darcian velocity and mass balance in the finite element modeling of groundwater flow[J]. Water Resources Research, 1981, 17(5): 1529-1534. doi: 10.1029/WR017i005p01529
[6]	D'ANGELO C, SCOTTI A. A mixed finite element method for Darcy flow in fractured porous media with non-matching grids[J]. ESAIM: Mathematical Modelling and Numerical Analysis, 2012, 46(2): 465-489. doi: 10.1051/m2an/2011148
[7]	赵文凤, 谢一凡, 吴吉春. 一种模拟节点达西渗透流速的双重网格多尺度有限单元法[J]. 岩土工程学报, 2020, 42(8): 1474-1481. https://www.cnki.com.cn/Article/CJFDTOTAL-YTGC202008018.htm ZHAO Wenfeng, XIE Yifan, WU Jichun. A dual-mesh multiscale finite element method for simulating nodal Darcy velocities in aquifers[J]. Chinese Journal of Geotechnical Engineering, 2020, 42(8): 1474-1481. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-YTGC202008018.htm
[8]	WILLIS R, YEH W W G. Groundwater systems planning and management[M]. Englewood Cliffs, NJ: Prentice-Hall, 1987
[9]	姚磊华. 遗传算法和高斯牛顿法联合反演地下水渗流模型参数[J]. 岩土工程学报, 2005, 27(8): 885-890. doi: 10.3321/j.issn:1000-4548.2005.08.007 YAO Leihua. Parameters identification of groundwater flow model with genetic algorithm and Gauss-Newton Method[J]. Chinese Journal of Geotechnical Engineering, 2005, 27(8): 885-890. (in Chinese) doi: 10.3321/j.issn:1000-4548.2005.08.007
[10]	PINKUS A. Approximation theory of the MLP model in neural networks[J]. Acta Numerica, 1999, 8: 143-195. doi: 10.1017/S0962492900002919
[11]	GUNES B A, PEARLMUTTER BARAK A, ANDREYEVICH R A, et al. Automatic differentiation in machine learning: a survey[J]. Journal of Machine Learning Research, 2018, 18(1): 5595-5637.
[12]	HAGHIGHAT E, RAISSI M, MOURE A, et al. A deep learning framework for solution and discovery in solid mechanics[EB/OL]. 2020: arXiv: 2003.02751. https://arxiv.org/abs/2003.02751
[13]	唐明健, 唐和生. 基于物理信息的深度学习求解矩形薄板力学正反问题[J]. 计算力学学报, 2022, 39(1): 120-128. https://www.cnki.com.cn/Article/CJFDTOTAL-JSJG202201018.htm TANG Mingjian, TANG Hesheng. A physics-informed deep learning method for solving forward and inverse mechanics problems of thin rectangular plates[J]. Chinese Journal of Computational Mechanics, 2022, 39(1): 120-128. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JSJG202201018.htm
[14]	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.
[15]	CHEN Z, LIU Y, SUN H. Physics-informed learning of governing equations from scarce data[J]. Nature Communications, 2021, 12: 6136.
[16]	ZHANG Q, CHEN Y L, YANG Z Y, et al. Multi-constitutive neural network for large deformation poromechanics problem[EB/OL]. 2020: arXiv: 2010.15549. https://arxiv.org/abs/2010.15549
[17]	BEKELE Y W. Physics-informed deep learning for flow and deformation in poroelastic media[EB/OL]. 2020: arXiv: 2010.15426. https://arxiv.org/abs/2010.15426
[18]	兰鹏, 李海潮, 叶新宇, 等. PINNs算法及其在岩土工程中的应用研究[J]. 岩土工程学报, 2021, 43(3): 586-592. https://www.cnki.com.cn/Article/CJFDTOTAL-YTGC202103028.htm LAN Peng, LI Haichao, YE Xinyu, et al. PINNs algorithm and its application in geotechnical engineering[J]. Chinese Journal of Geotechnical Engineering, 2021, 43(3): 586-592. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-YTGC202103028.htm
[19]	HE Q Z, BARAJAS-SOLANO D, TARTAKOVSKY G, et al. Physics-informed neural networks for multiphysics data assimilation with application to subsurface transport[J]. Advances in Water Resources, 2020, 141: 103610.
[20]	TARTAKOVSKY A M, MARRERO C O, PERDIKARIS P, et al. Physics-informed deep neural networks for learning parameters and constitutive relationships in subsurface flow problems[J]. Water Resources Research, 2020, 56(5): e2019WR026731.
[21]	BANDAI T, GHEZZEHEI T A. Physics-informed neural networks with monotonicity constraints for Richardson- richards equation: estimation of constitutive relationships and soil water flux density from volumetric water content measurements[J]. Water Resources Research, 2021, 57(2): e2020WR027642.
[22]	BANDAI T, GHEZZEHEI T A. Forward and inverse modeling of water flow in unsaturated soils with discontinuous hydraulic conductivities using physics-informed neural networks with domain decomposition[J]. Hydrology and Earth System Sciences, 2022, 26(16): 4469-4495.
[23]	LU L, MENG X H, MAO Z P, et al. DeepXDE: a deep learning library for solving differential equations[J]. SIAM Review, 2021, 63(1): 208-228.
[24]	SUN L N, GAO H, PAN S W, et al. Surrogate modeling for fluid flows based on physics-constrained deep learning without simulation data[J]. Computer Methods in Applied Mechanics and Engineering, 2020, 361: 112732.
[25]	LAGARI P L, TSOUKALAS L H, SAFARKHANI S, et al. Systematic construction of neural forms for solving partial differential equations inside rectangular domains, subject to initial, boundary and interface conditions[J]. International Journal on Artificial Intelligence Tools, 2020, 29(5): 2050009.
[26]	LU L, PESTOURIE R, YAO W J, et al. Physics-informed neural networks with hard constraints for inverse design[J]. SIAM Journal on Scientific Computing, 2021, 43(6): B1105-B1132.
[27]	陆至彬, 瞿景辉, 刘桦, 等. 基于物理信息神经网络的传热过程物理场代理模型的构建[J]. 化工学报, 2021, 72(3): 1496-1503. https://www.cnki.com.cn/Article/CJFDTOTAL-HGSZ202103031.htm LU Zhibin, QU Jinghui, LIU Hua, et al. Surrogate modeling for physical fields of heat transfer processes based on physics-informed neural network[J]. CIESC Journal, 2021, 72(3): 1496-1503. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-HGSZ202103031.htm
[28]	ABADI M, BARHAM P, CHEN J M, et al. TensorFlow: a system for large-scale machine learning[C]//Proceedings of the 12th USENIX conference on Operating Systems Design and Implementation. November 2 - 4, 2016, Savannah, GA, USA. New York: ACM, 2016: 265-283.
[29]	ADAM P, GROSS S, CHINTALA S, et al. Automatic differentiation in PyTorch[C]// 31st Conference on Neural Information Processing Systems (NIPS 2017), 2017, Long Beach.
[30]	MCKAY M D, BECKMAN R J, CONOVER W J. A comparison of three methods for selecting values of input variables in the analysis of output from a computer code[J]. Technometrics, 2000, 42(1): 55-61.
[31]	KINGMA D P, BA J. Adam: A method for stochastic optimization[C]// 3rd International Conference on Learning Representations, 2015, San Diego.
[32]	ZHU C, BYRD R H, LU P, et al. Algorithm 778: L-BFGS-B: Fortran subroutines for large-scale bound-constrained optimization[J]. ACM Transactions on Mathematical Software, 1997, 23(4): 550–560.