Learning complex nonlinear physical systems using wavelet neural operators
Yanan Guo(郭亚楠)1,2,3, Xiaoqun Cao(曹小群)1,†, Hongze Leng(冷洪泽)1, and Junqiang Song(宋君强)1
1 College of Meteorology and Oceanography, National University of Defense Technology, Changsha 410073, China; 2 College of Computer, National University of Defense Technology, Changsha 410073, China; 3 Naval Aviation University, Huludao 125001, China
Abstract Nonlinear science is a fundamental area of physics research that investigates complex dynamical systems which are often characterized by high sensitivity and nonlinear behaviors. Numerical simulations play a pivotal role in nonlinear science, serving as a critical tool for revealing the underlying principles governing these systems. In addition, they play a crucial role in accelerating progress across various fields, such as climate modeling, weather forecasting, and fluid dynamics. However, their high computational cost limits their application in high-precision or long-duration simulations. In this study, we propose a novel data-driven approach for simulating complex physical systems, particularly turbulent phenomena. Specifically, we develop an efficient surrogate model based on the wavelet neural operator (WNO). Experimental results demonstrate that the enhanced WNO model can accurately simulate small-scale turbulent flows while using lower computational costs. In simulations of complex physical fields, the improved WNO model outperforms established deep learning models, such as U-Net, ResNet, and the Fourier neural operator (FNO), in terms of accuracy. Notably, the improved WNO model exhibits exceptional generalization capabilities, maintaining stable performance across a wide range of initial conditions and high-resolution scenarios without retraining. This study highlights the significant potential of the enhanced WNO model for simulating complex physical systems, providing strong evidence to support the development of more efficient, scalable, and high-precision simulation techniques.
Fund: Project supported by the National Natural Science Foundation of China (Grant Nos. 42005003 and 41475094).
Corresponding Authors:
Xiaoqun Cao
E-mail: caoxiaoqun@nudt.edu.cn
Cite this article:
Yanan Guo(郭亚楠), Xiaoqun Cao(曹小群), Hongze Leng(冷洪泽), and Junqiang Song(宋君强) Learning complex nonlinear physical systems using wavelet neural operators 2025 Chin. Phys. B 34 034702
[1] Kajishima T and Taira K 2016 Computational fluid dynamics: incompressible turbulent flows (Springer) [2] Piquet J 2013 Turbulent flows: models and physics (Springer Science & Business Media) [3] Argyropoulos C D and Markatos N 2015 Applied Mathematical Modelling 39 693 [4] Kadaverugu R, Sharma A, Matli C and Biniwale R 2019 Asia-Pacific Journal of Atmospheric Sciences 55 539 [5] Van Heerwaarden C C, Van Stratum B J, Heus T, Gibbs J A, Fedorovich E and Mellado J P 2017 Geoscientific Model Development 10 3145 [6] Chung D and Pullin D 2010 Journal of Fluid Mechanics 643 279 [7] Sharfuddin A A M and Ladeinde F 2023 AIAA Aviation 2023 Forum p. 4415 [8] Ishihara T, Gotoh T and Kaneda Y 2009 Annual Review of Fluid Mechanics 41 165 [9] Fox R O 2012 Annual Review of Fluid Mechanics 44 47 [10] Lee M and Moser R D 2015 Journal of Fluid Mechanics 774 395 [11] Tryggvason G, Scardovelli R and Zaleski S 2011 Direct numerical simulations of gas-liquid multiphase flows (Cambridge university press) [12] Panchigar D, Kar K, Shukla S, Mathew R M, Chadha U and Selvaraj S K 2022 Neural Computing and Applications 34 21677 [13] Guo Y, Cao X, Song J and Leng H 2024 Chin. Phys. B 33 020203 [14] Guo Y, Cao X and Peng K 2023 Commun. Theor. Phys. 75 095003 [15] Guo Y, Cao X, Liu B and Gao M 2020 Applied Sciences 10 5917 [16] Kou J and Zhang W 2021 Progress in Aerospace Sciences 125 100725 [17] Ladický L, Jeong S, Solenthaler B, Pollefeys M and Gross M 2015 ACM Transactions on Graphics (TOG) 34 1 [18] Mendez M A, Ianiro A, Noack B R and Brunton S L 2023 Datadriven fluid mechanics: combining first principles and machine learning (Cambridge University Press) [19] Kochkov D, Smith J A, Alieva A, Wang Q, Brenner M P and Hoyer S 2021 Proc. Natl. Acad. Sci. USA 118 e2101784118 [20] Guo Y, Cao X, Song J, Leng H and Peng K 2023 Physics of Fluids 35 106603 [21] Sousa P, Rodrigues C V and Afonso A 2024 Computer Methods in Applied Mechanics and Engineering 429 117133 [22] Guo Y, Cao X, Zhou M, Leng H and Song J 2024 Physics of Fluids 36 115110 [23] Kovachki N, Li Z, Liu B, Azizzadenesheli K, Bhattacharya K, Stuart A and Anandkumar A 2023 Journal of Machine Learning Research 24 1 [24] Azizzadenesheli K, Kovachki N, Li Z, Liu-Schiaffini M, Kossaifi J and Anandkumar A 2024 Nature Reviews Physics 6 320 [25] Li Z, Huang D Z, Liu B and Anandkumar A 2023 Journal of Machine Learning Research 24 1 [26] Kovachki N, Lanthaler S and Mishra S 2021 Journal of Machine Learning Research 22 1 [27] Zhang T, Li L, López-Marrero V, Lin M, Liu Y, Yang F, Yu K and Atif M 2024 Journal of Advances in Modeling Earth Systems 16 e2023MS003898 [28] Cao Q, Goswami S and Karniadakis G E 2024 Nature Machine Intelligence 6 631 [29] Shih B, Peyvan A, Zhang Z and Karniadakis G E 2025 Computer Methods in Applied Mechanics and Engineering 434 117560 [30] Niarchos V and Papageorgakis C 2024 Phys. Rev. D 110 045020 [31] Tripura T and Chakraborty S 2023 Computer Methods in Applied Mechanics and Engineering 404 115783 [32] Tripura T, Awasthi A, Roy S and Chakraborty S 2023 Computer Methods and Programs in Biomedicine 232 107436 [33] Lu C, Liu Y and Niu S 2011 Journal of Geophysical Research: Atmospheres 116 3740 [34] Lu C, Liu Y, Niu S, Krueger S andWagner T 2013 Journal of Geophysical Research: Atmospheres 118 185 [35] Gao Z, Liu Y, Li X and Lu C 2018 Journal of Geophysical Research: Atmospheres 123 2194 [36] Schneiders L, Meinke M and Schröder W 2017 Journal of Fluid Mechanics 819 188
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.
