| COMPUTATIONAL PROGRAMS FOR PHYSICS |
Prev
Next
|
|
|
Improved physics-informed neural networks incorporating lattice Boltzmann method optimized by tanh robust weight initialization |
| Chenghui Yang(杨程晖)1, Minglei Shan(单鸣雷)1,†, Mengyu Feng(冯梦宇)1, Ling Kuai(蒯玲)1, Yu Yang(杨雨)2, Cheng Yin(殷澄)1, and Qingbang Han(韩庆邦)1 |
1 College of Information Science and Engineering, Hohai University, Changzhou 213200, China;
2 College of Information Science and Technology, Nanjing Forestry University, Nanjing 210037, China |
|
|
|
|
Abstract Physics-informed neural networks (PINNs) have shown considerable promise for performing numerical simulations in fluid mechanics. They provide mesh-free, end-to-end approaches by embedding physical laws into their loss functions. However, when addressing complex flow problems, PINNs still face some challenges such as activation saturation and vanishing gradients in deep network training, leading to slow convergence and insufficient prediction accuracy. We present physics-informed neural networks incorporating lattice Boltzmann method optimized by tanh robust weight initialization (T-PINN-LBM) to address these challenges. This approach fuses the mesoscopic lattice Boltzmann model with the automatic differentiation framework of PINNs. It also implements a tanh robust weight initialization method derived from fixed point analysis. This model effectively mitigates activation and gradient decay in deep networks, improving convergence speed and data efficiency in multiscale flow simulations. We validate the effectiveness of the model on the classical arithmetic example of lid-driven cavity flow. Compared to the traditional Xavier initialized PINN and PINN-LBM, T-PINNLBM reduces the mean absolute error (MAE) by one order of magnitude at the same network depth and maintains stable convergence in deeper networks. The results demonstrate that this model can accurately capture complex flow structures without prior data, providing a new feasible pathway for data-free driven fluid simulation.
|
Received: 18 June 2025
Revised: 22 July 2025
Accepted manuscript online: 18 August 2025
|
|
PACS:
|
07.05.Mh
|
(Neural networks, fuzzy logic, artificial intelligence)
|
| |
02.60.Cb
|
(Numerical simulation; solution of equations)
|
| |
02.30.Jr
|
(Partial differential equations)
|
| |
84.35.+i
|
(Neural networks)
|
|
| Fund: Project supported by the National Natural Science Foundation of China (Grant Nos. 12474453, 12174085, and 12404530). |
Corresponding Authors:
Minglei Shan
E-mail: shanml@hhu.edu.cn
|
Cite this article:
Chenghui Yang(杨程晖), Minglei Shan(单鸣雷), Mengyu Feng(冯梦宇), Ling Kuai(蒯玲), Yu Yang(杨雨), Cheng Yin(殷澄), and Qingbang Han(韩庆邦) Improved physics-informed neural networks incorporating lattice Boltzmann method optimized by tanh robust weight initialization 2025 Chin. Phys. B 34 110701
|
[1] Zienkiewicz O C and Taylor R L 2013 Finite Element Method (Oxford: Amsterdam Butterworth-Heinemann Press) pp. 705–714
[2] Liu Y Q, Cheng R J and Ge H X 2013 Chin. Phys. B 22 100204
[3] Cardiff P and Demirdzic I 2021 Arch. Comput. Meth. Engin. 28 3721
[4] Raissi M, Perdikaris P and Karniadakis G E 2018 J. Comput. Phys. 378 686
[5] Mao Z P, Jagtap A D and Karniadakis G E 2020 Comput. Meth. Appl. Mech. Engin. 360 112789
[6] Shukla K, Leoni D and Karniadakis G E 2020 J. Nondest. Eval. 39 61
[7] Lu L, Jin P Z, Pang G F, Zhang Z Q and Karniadakis G E 2021 Nat. Mach. Intell. 3 218
[8] Karniadakis G E, Kevrekidis I G, Lu L, Perdikaris P, Wang S F and Yang L 2021 Nat. Rev. Phys. 3 422
[9] Yin M L, Zheng X N, Humphrey J and Karniadakis G E 2021 Comput. Meth. Appl. Mech. Engin. 375 113603
[10] Bararnia H and Esmaeilpour M 2022 Inter. Com. Heat. Mass Tran. 132 105890
[11] Cuomo S, Di C, Vincenzo S, Giampaolo F, Rozza G, Raissi M and Piccialli F 2022 J. Sci. Comput. 92 88
[12] Hanna J M, Aguado J, Askri R and Borzacchiello D 2022 Comput. Meth. Appl. Mech. Engin. 396 115100[13] Hosseini V R, Mehrizi A, Gungor A and Afrouzi H 2023 Fuel 332 125908
[14] Zhu J A, Xue Y H and Liu Z S 2024 Appl. Math. Mech. 45 1685
[15] Liu J N, Hou Q Z,Wei J G and Sun ZW2023 Chin. Phys. B 32 070702
[16] ZhaWS, Chen D S, Li D L, Shen L H and Chen E Y 2025 Chin. Phys. B 34 040701
[17] Chen S Y and Doolen G D 2003 Annu. Rev. Flu. Mech. 30 329
[18] Lou Q, Meng X H and Karniadakis G E 2021 J. Comput. Phys. 447 110676
[19] Liu Z X, Chen Y, Song G, Song W and Xu J 2023 Mathematics 11 4147
[20] Fang D H, Tan J F and Xu J 2022 Oce. Engin. 263 112360
[21] Qian Y H, Humieres D and Lallemand P 1992 Europhys. Lett. 17 479
[22] He X Y and Luo L S 1997 Phys. Rev. E 55 R6333
[23] Luo L S and Girimaji S 2003 Phys. Rev. E. 67 036302
[24] Boon J P 2003 Euro. J. Mech. Fluids. 22 101
[25] Lee H, Choi H and Kim H 2025 arXiv:2410.02242 [cs.LG]
[26] Kingma D and Ba J 2014 arXiv:1412.6980v9 [cs.LG]
[27] Li Y F, Shi Q Z, Li Y, Song X J, Liu C C and Wang W Q 2021 Chin. Phys. B 30 014302
[28] Wang S F, Teng Y J and Perdikaris P 2021 SIAM J. Sci. Comput. 43 A3055
[29] Chiu P,Wong J C, Ooi C, DaoMand Ong Y 2022 Comput. Meth. Appl. Mech. Engin. 395 114909 |
| No Suggested Reading articles found! |
|
|
Viewed |
|
|
|
Full text
|
|
|
|
|
Abstract
|
|
|
|
|
Cited |
|
|
|
|
Altmetric
|
|
blogs
Facebook pages
Wikipedia page
Google+ users
|
Online attention
Altmetric calculates a score based on the online attention an article receives. Each coloured thread in the circle represents a different type of online attention. The number in the centre is the Altmetric score. Social media and mainstream news media are the main sources that calculate the score. Reference managers such as Mendeley are also tracked but do not contribute to the score. Older articles often score higher because they have had more time to get noticed. To account for this, Altmetric has included the context data for other articles of a similar age.
View more on Altmetrics
|
|
|
