Skip to content
Open access

Soliton solutions to KdV and KP equations using physics-informed neural networks

2026 · Journal of Nonlinear Waves · 0 citations · 30 references

TL;DR

This work systematically investigates the ability of PINNs to capture the nonlinear dynamics of the Korteweg–de Vries (KdV) equation and its two-dimensional generalization, the Kadomtsev–Petviashvili (KP) equation, and compares network architectures, activation functions and optimization strategies while benchmarking PINN performance against analytic solutions.

Abstract

Partial differential equations (PDEs) model a wide range of physical phenomena but are often difficult to solve analytically or numerically, particularly when they are nonlinear and contain high-order derivative terms, as in water-wave problems. Although many numerical methods exist, they are typically tailored to specific equations. Physics-informed neural networks (PINNs) have recently emerged as a flexible alternative by incorporating the governing PDE, together with initial and boundary conditions, directly into the network’s loss function. In this work, we systematically investigate the ability of PINNs to capture the nonlinear dynamics of the Korteweg–de Vries (KdV) equation and its two-dimensional generalization, the Kadomtsev–Petviashvili (KP) equation. Recovering nonlinear soliton interactions from the governing PDE alone therefore provides a stringent test of whether the network has learned the underlying dynamics rather than simply reproducing training data. We examine increasingly complex interacting-soliton configurations for KdV and analogous line-soliton and junction solutions for KP, evaluating how accuracy, training time and network complexity scale with problem difficulty. We also compare network architectures, activation functions and optimization strategies while benchmarking PINN performance against analytic solutions.

Read PDF

Similar papers

Oct 2026

Learning approximate solutions of partial differential equations within the physics-informed framework

Variants of Physics-informed neural networks (PINNs) often treat neural networks as the main solution representation, yet solutions are encoded implicitly in high-dimensional weights, making training behavior and outcomes difficult to predict or analyze across different partial differential equation (PDE) settings. In...

Kai-Wei Cong, Shuang-Qing Chen, Bing Guan · 0 citations
Open access Oct 2026

Analytical soliton solutions and physics-informed neural network validation for the generalized higher-order nonlinear Schrödinger equation

This paper examines the generalized higher-order nonlinear Schrödinger equation (GHONLSE), which is an underlying model used to investigate the wave propagation in nonlinear dispersive media such as optical fibers and plasma systems, in which the higher-order dispersion and nonlinear interaction effects are subject to...

Sabah Iqbal, Faisal Yasin, A. Msmali et al. · 0 citations
Open access Sep 2026

Physics-Informed Deep Learning for Soliton Dynamics and Energy Flow Control in Nonlinear Physical Systems

Nonlinear field equations that model high-energy physical systems are frequently hampered by limited integration of deep learning frameworks and the absence of robust hybrid analytical-computational methodologies. In the current research, we overcome such challenges by examining the complex coupled Higgs field equati...

Sajjad Ahmad, S. Hussain, Muhammad Shoaib Saleem et al. · 0 citations
Open access Sep 2026

Scientific machine learning meets semi-analytical computation: a hybrid NIM-PINN approach for nonlinear PDEs

In this paper, a hybrid semi-analytical–deep learning framework for the solution of nonlinear partial differential equations (PDEs) is proposed that integrates the New Iterative Method (NIM) with Physics-Informed Neural Networks (PINNs). Low–order NIM expansion provides an analytic baseline that meets the required init...

A. Alshehry, Saima Noor, Humaira Yasmin et al. · 0 citations
Open access 2026

A Physics-Informed Neural Network Approach for Solving the Monge-Ampère Partial Differential Equation

This paper presents a comparative study of classical numerical and machine learning techniques for solving the nonlinear Monge–Ampère Equation (MAE). A finite difference method (FDM) implemented in MATLAB is first validated using a quadratic exact solution on a square domain. Two neural network approaches are then inve...

C. Ogbonna, Maliki Olaniyi S, Nkem Ogbonna · 0 citations
Preprint Sep 2026

Towards solving General Relativity with Physics-Informed Neural Networks

Physics-Informed Neural Networks (PINNs) are a machine-learning framework for approximating solutions to systems of partial differential equations by constraining neural networks to satisfy the underlying physical laws. The resulting continuous representation does not require a predefined computational mesh and can be...

J. A. Carretero, Jorge F. Urbán, F. Abalos et al. · 3 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.