Research on highly nonlinear Richards infiltration model based on PINNs

Abstract

Abstract: For the Richards infiltration model of unsaturated soil with highly nonlinear coefficients, the physics-informed neural networks(PINNs) is applied to solve it and the network prediction results are verified by the finite difference method. It is found that the prediction results of PINNs are basically consistent with the prediction results of finite difference method. The influence of hyperparameter on the error of PINNs is further investigated to determine the impact of factors such as training set size and number of network layers on PINNs training set and test set errors. With reasonable hyperparameter adjustment, the PINNs prediction model exhibits good training performance in the highly nonlinear infiltration model. This computational method can be widely applied to solve airport engineering problems such as heat conduction, water vapor migration and stress balance.

Key words: highly nonlinear coefficient, infiltration model, physics -informed neural networks (PINNs), finite differ鄄 ence method, hyperparameter adjustment

CLC Number:

HUO Haifeng, HUANG Haoyu, LI Qiang, HU Biao, ZHANG Zhaowen. Research on highly nonlinear Richards infiltration model based on PINNs[J]. Journal of Civil Aviation University of China, 2023, 41(5): 6-12.

References 22

[1]	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.
[2]	李峻屹. 基于 PINNs 方法求解非定常 Stokes 方程[J]. 陕西科技大学学报, 2021, 39(3): 182-186.
[3]	RAO C P, SUN H, LIU Y. Physics -informed deeplearning for incompressible laminar flows[J]. Theoretical and Applied Mechanics Letters, 2020, 10(3): 207-212.
[4]	MAO Z P, JAGTAP A D, KARNIADAKIS G E. Physics -informed neural networks for high-speed flows[J]. Computer Methods in Applied Mechanics and Engineering, 2020, 360: 112789.
[5]	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): 208228.
[6]	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.
[7]	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.
[8]	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.
[9]	RYCK T D, MISHRA S. Error analysis for physics informed neural net－ works (PINNs) approximating Kolmogorov PDEs[EB/OL].(2021-06-28) [2022-10-07].2021:arXiv:2106. 14473. https://arxiv.org/abs/2106. 14473.
[10]	BAI G M, KOLEY U, MISHRA S, et al Physics informed neural net－ works (PINNs) for approximating nonlinear dispersive PDEs [EB/OL] .(2021-04 -12) [2022 -10 -07]. 2021: arXiv:2104.05584. https://arxiv.org/abs/2104. 05584.
[11]	RASHT -BEHESHT M, HUBER C, SHUKLA K, et al Physics -in－ formed neural networks (PINNs) for wave propagation and full waveform inversions[J]. Journal of Geophysical Research: Solid Earth, 2022, 127 (5): e2021JB023120.
[12]	PANG G F, LU L, KARNIADAKIS G E. fPINNs: fractional physicsinformed neural networks[J]. SIAM Journal on Scientific Computing, 2019, 41(4): A2603-A2626.
[13]	WANG S P, ZHANG H, JIANG X Y. Fractional physics-informed neu－ ral networks for time-fractional phase field models[J]. Nonlinear Dy－ namics, 2022, 110(3): 2715-2739.
[14]	GUO L, WU H, YU X C, et al Monte Carlo fPINNs: deep learning method for forward and inverse problems involving high dimensional fractional partial differential equations[J]. Computer Methods in Ap－ plied Mechanics and Engineering, 2022, 400: 115523.
[15]	YUAN L, NI Y Q, DENG X Y, et al A-PINN: auxiliary physics-in－ formed neural networks for forward and inverse problems of nonlinear integro-differential equations[J]. Journal of Computational Physics, 2022, 462: 111260.
[16]	MUALEM Y. A new model for predicting the hydraulic conductivity of unsaturated porous media[J]. Water Resources Research, 1976, 12(3): 513-522.
[17]	BITTELLI M, CAMPBELL G S, TOMEI F. Soil physics with Python: transport in the soil-plant-atmosphere system[M]. NY:OUP Oxford, 2015.
[18]	ABADI M, BARHAM P, CHEN J M, et al TensorFlow: a system for large-scale machine learning [EB/OL]（． 2016-05-27）[2022-10-17].2016: arXiv: 1605.08695．https：//arxiv．org/abs/1605.08695．pdf.
[19]	KINGMA D P, BA J． Adam: a method for stochastic optimization[EB/ OL]（． 2014-12-22）[2022-10-17]. 2014: arXiv：1412.6980．https：//arx－ iv．org/abs/1412.6980．pdf.
[20]	王俊智, 梁四海, 万力, 等. 高寒地区冻土活动层温度场数值模拟[J]. 人民黄河, 2013, 35(3): 20-23.
[21]	唐国航, 王乃东, 姚仰平. 基于 PDV 模型对一维“锅盖效应”气态水迁移的研究[J]. 岩土工程学报, 2022, 44(S2): 164-168.
[22]	张召峰, 傅鹤林, 李成威, 等. 移动飞机荷载对机场跑道下方土体的附加应力分析[J]. 现代隧道技术, 2022, 59(S1): 1111-1119.