Hostname: page-component-7b7567c8c-fvjsw Total loading time: 0 Render date: 2026-11-17T05:42:35.930Z Has data issue: false hasContentIssue false

Convergent series of Stokes waves of arbitrary height in deep water via machine learning

Published online by Cambridge University Press:  16 December 2025

Chong Lin
Affiliation:
School of Ocean and Civil Engineering, Shanghai Jiao Tong University, Shanghai 200240, PR China
Shijun Liao*
Affiliation:
School of Ocean and Civil Engineering, Shanghai Jiao Tong University, Shanghai 200240, PR ChinaState Key Laboratory of Ocean Engineering, Shanghai 200240, PR China
*
Corresponding author: Shijun Liao, sjliao@sjtu.edu.cn

Abstract

Permanent gravity waves propagating in deep water, spanning amplitudes from infinitesimal to their theoretical limiting values, remain a classical yet challenging problem due to its inherent nonlinear complexities. Traditional analytical and numerical methods encounter substantial difficulties near the limiting wave condition due to singularities at sharp wave crests. In this study, we propose a novel hybrid framework combining the homotopy analysis method (HAM) with machine learning (ML) to efficiently compute convergent series solutions of Stokes waves in deep water for arbitrary wave amplitudes from small to theoretical limiting values, which show excellent agreement with established benchmarks. We introduce a neural network trained using only 20 representative cases whose series solution are given by means of HAM, which can rapidly predict series solutions across arbitrary steepness levels, substantially improving computational efficiency. Additionally, we develop a neural network to gain the inverse mapping from the conformal coordinates $(\theta , r)$ to the physical coordinates $(x,y)$, facilitating explicit and intuitive representations of series solutions in physical plane. This HAM–ML hybrid framework represents a powerful and efficient approach to compute convergent series in a whole range of physical parameters for water waves with arbitrary wave height including even limiting waves. In this way we establish a new paradigm to quickly obtain convergent series solutions of complex nonlinear systems for a whole range of physical parameters, thereby significantly broadening the scope of series solutions that can be easily gained by means of HAM even for highly nonlinear problems in science and engineering.

Information

Type
JFM Papers
Copyright
© The Author(s), 2025. Published by Cambridge University Press
Figure 0

Figure 1.(a) The $z$ plane, (b) the $\zeta$ plane.

Figure 1

Figure 2.Convergence behaviour of selected Fourier coefficients $ a_1$, $ a_5$, $ a_{10}$, $ a_{15}$, $ a_{20}$, $ a_{25}$, $ a_{30}$, and the Bernoulli constant $ K$ for a Stokes wave in infinite depth with steepness $ A/L = 0.07$, based on 1000th-order HAM approximations. The y-axis is the successive order difference $|a_{j,n+1}-a_{j,n}|$ for coefficient $a_{\kern-1pt j}$, and the x-axis is the HAM order $n$.

Figure 2

Table 1.Comparison of the squared wave velocity $ c^2$ and the Bernoulli constant $ K$ for Stokes waves of various steepness values in infinite depth water between the present HAM-based solutions and the reference data reported by Schwartz (1974).

Figure 3

Figure 3.Stokes wave profiles in infinite depth for different wave steepness values.

Figure 4

Figure 4.Comparison of limiting Stokes wave profiles in infinite depth water obtained by different methods.

Figure 5

Figure 5.Stokes wave flow fields of varying steepness in infinite-depth water. Arrows represent the local velocity vectors, while the colour contours depict the magnitude of the flow speed. Here (a) $A/L=0.040$; (b) $A/L=0.100$; (c) $A/L=0.130$; (d) limiting wave.

Figure 6

Figure 6.Schematic of ML model used to predict the solution of the wave in infinite depth water with arbitrary wave height. The corresponding ML codes can be freely downloaded from the website https://github.com/sjtu-liao/deep-water-Stokes-wave and https://numericaltank.sjtu.edu.cn/NonlinearWaves.htm (see ‘Convergent Series of Stokes Wave of Arbitrary Height in Deep Water’).

Figure 7

Figure 7.Here (a) ML, $A/L=0.070$; (b) ML, $A/L=0.120$; (c) ML, limiting wave; (d) HAM, $A/L=0.070$; (e) HAM, $A/L=0.120$; (f) HAM, limiting wave. Comparison of wave flow fields using the ML solutions and the benchmark HAM solutions for wave steepness values of $A/L = 0.07$, $0.12$ and the limiting case in infinite-depth water. Panels (a)–(c) show the velocity fields obtained via the ML model, while panels (d)–(f) present the corresponding HAM results. Arrows represent the local velocity directions, and colour contours indicate velocity magnitudes.

Figure 8

Table 2.Mean absolute errors of the ML solutions for various physical quantities corresponding to different wave steepness values in infinite-depth water.

Figure 9

Figure 8.Comparison of free-surface profiles using the ML solutions and the benchmark HAM solutions for different wave steepness. Here lines, ML solutions; symbols, HAM solutions.

Figure 10

Figure 9.Schematic of the ML framework for inverse coordinate transformation. The dashed box indicates the wave model introduced in § 3.2, while the dash–dotted box denotes the neural network designed for performing the inverse transformation. The corresponding ML codes can be freely downloaded on the website https://github.com/sjtu-liao/deep-water-Stokes-wave and https://numericaltank.sjtu.edu.cn/NonlinearWaves.htm (see ‘Convergent Series of Stokes Wave of Arbitrary Height in Deep Water’).

Figure 11

Table 3.Mean absolute errors of the predicted velocity fields obtained from the inverse coordinate transformation neural network, evaluated at selected spatial locations across a range of wave steepness values.

Figure 12

Figure 10.(Here (a) $\zeta$ plane, $A/L=0.060$; (b) $z$ plane, $A/L=0.060$; (c) error, $A/L=0.060$; (d) $\zeta$ plane, $A/L=0.110$; (e) $z$ plane, $A/L=0.110$; (f) error, $A/L=0.110$; (g) $\zeta$ plane, $A/L=0.135$; (h) $z$ plane, $A/L=0.135$; (i) error, $A/L=0.135$; (j) $\zeta$ plane, limiting wave; (k) $z$ plane, limiting wave; (l) error, limiting wave. Comparison of flow fields and corresponding error distributions for selected wave steepness values before and after applying the neural-network-based inverse coordinate transformation. For each case, panels (a), (d), (g), (j) and panels (b), (e), (h), (k) display the velocity fields obtained from the HAM solution and those reconstructed via neural network inversion, respectively, while panels (c), (f), (i), (l) show the corresponding pointwise error in velocity magnitude, defined as $\varDelta |v| = |v_{{NN}} - v_{\textit{HAM}}|$. Slightly elevated errors are observed in localized regions with strong nonlinear features or relatively limited spatial resolution, particularly near the wave crest and regions with steep velocity gradients.

Figure 13

Figure 11.Comparison of free-surface profiles using the ML solutions and the fifth-order Stokes expansion solutions for different wave steepness values.