Abstract:The Burgers hierarchy consists of nonlinear evolutionary PDEs with increasingly higher-order dispersive and nonlinear terms, resulting in a rich variety of soliton solutions. Notable members include the Burgers equation and the Sharma-Tasso-Olver equation, which are applied in plasma physics, fluid mechanics, optics, and biophysics to describe nonlinear waves in inhomogeneous media. Various soliton and multi-soliton solutions have been identified, and the fission and fusion of solitons have been explored through analytical and numerical techniques. Recently, deep learning methods, particularly Physics Informed Neural Networks (PINNs), have emerged as powerful tools for solving PDEs. These methods use deep neural networks to minimize PDE residuals and satisfy boundary and initial conditions while fitting relevant data. Although PINNs have been applied to equations like Burgers' and Korteweg-de Vries, higher-order members of the Burgers hierarchy remain unexplored. In this study, we employ a PINN algorithm to approximate multi-soliton solutions of linear combinations of equations within the Burgers hierarchy. This semi-supervised approach encodes the PDE and relevant data, determining PDE parameters and resolving the linear combination to discover the PDE that best describes the data. Additionally, we employ gradient-enhanced PINNs (gPINNs) and a conservation law specific to the generic Burgers' hierarchy to improve training accuracy. The results demonstrate the effectiveness of PINNs in describing multi-soliton solutions within the generic Burgers' hierarchy. They also verify the potential for training refinement and accuracy improvement using enhanced approaches in certain cases, while enabling the inference of the PDE model that best describes the observed solitary structures.

What problem does this paper attempt to address?

### Problems Addressed by the Paper This paper aims to solve the multi-soliton solution problem in high-order Burgers hierarchy equations using Physics Informed Neural Networks (PINNs). Specifically, the paper focuses on the following aspects: 1. **Approximation of Multi-Soliton Solutions**: - Researchers constructed a general Burgers hierarchy equation by linearly combining different members of the Burgers hierarchy equations. - The PINNs algorithm is used to approximate the multi-soliton solutions of these equations, particularly the three-soliton solution. 2. **Data-Driven Equation Discovery**: - By using a semi-supervised learning approach, combining experimental data and model equations, the PDE parameters are determined to discover the PDE model that best describes the data. - Especially for high-order Burgers hierarchy equations, such as the Sharma-Tasso-Olver (STO) equation and the fourth-order Burgers equation, existing research has not yet addressed these high-order members. 3. **Enhancing Training Accuracy**: - Introducing gradient-enhanced PINNs (gPINNs) and conservation law-constrained PINNs (cl-PINNs) to improve training accuracy. - By embedding the gradient information of the PDE residuals and conservation laws into the loss function, the training effect is improved. ### Main Contributions 1. **Approximation of Multi-Soliton Solutions**: - Successfully used the PINNs algorithm to approximate the multi-soliton solutions of high-order Burgers hierarchy equations, including the phenomena of three-soliton fusion and splitting. - The accuracy of these solutions was verified through experimental data. 2. **Data-Driven Equation Discovery**: - Successfully inferred the parameters of the PDE model from data using a semi-supervised learning approach, thereby determining the mathematical model that best describes the observed solitary structures. - Verified the potential of this method in identifying and quantifying physical mechanisms, such as the presence of dispersion and diffusion. 3. **Enhancing Training Accuracy**: - Significantly improved training accuracy by introducing gradient enhancement and conservation law constraints. - In some cases, these enhancement methods can further optimize training results, especially with larger learning rates. ### Experimental Results - **Three-Soliton Fusion**: - Successfully simulated the fusion process of three solitons at specific time and space points using the semi-supervised PINNs algorithm. - The difference between the training results and the exact solution is on the order of 10^-2, indicating high model accuracy. - **Three-Soliton Splitting**: - Similarly, using the semi-supervised PINNs algorithm, successfully simulated the process of a single solitary wave splitting into three solitons. - The training results are also highly consistent with the exact solution, verifying the effectiveness of the method. - **Effectiveness of Enhancement Methods**: - Significantly improved training accuracy by introducing gradient enhancement and conservation law constraints, especially with larger learning rates. - These enhancement methods can further optimize training results in some cases, but their advantages may diminish or reverse with smaller or dynamically adjusted learning rates. In summary, this paper successfully addresses the multi-soliton solution problem in high-order Burgers hierarchy equations by combining PINNs, gPINNs, and cl-PINNs methods, and demonstrates the great potential of data-driven methods in discovering and validating physical models.

Multi-soliton solutions and data-driven discovery of higher-order Burgers' hierarchy equations with physics informed neural networks

Approximate solution of KdV-Burgers equation using improved PINNs algorithm

Distributed physics informed neural network for data-efficient solution to partial differential equations

Burgers' pinns with implicit euler transfer learning

A Combination of Deep Neural Networks and Physics to Solve the Inverse Problem of Burger's Equation

Solving Viscous Burgers' Equation: Hybrid Approach Combining Boundary Layer Theory and Physics-Informed Neural Networks

Data-driven solutions and parameter estimations of a family of higher-order KdV equations based on physics informed neural networks

A two-stage physics-informed neural network method based on conserved quantities and applications in localized wave solutions

Data-driven solitons dynamics and parameters discovery in the generalized nonlinear dispersive mKdV-type equation via deep neural networks learning

Solving Partial Differential Equations Using Deep Learning and Physical Constraints

Explorations of certain nonlinear waves of the Boussinesq and Camassa–Holm equations using physics-informed neural networks

Data-driven soliton solutions and parameters discovery of the coupled nonlinear wave equations via a deep learning method

Binary structured physics-informed neural networks for solving equations with rapidly changing solutions

Predicting positon solutions of a family of Nonlinear Schrödinger equations through Deep Learning algorithm

Investigating the Ability of PINNs To Solve Burgers' PDE Near Finite-Time BlowUp

Operator Learning Enhanced Physics-informed Neural Networks for Solving Partial Differential Equations Characterized by Sharp Solutions

Solving Second-Order Nonlinear Evolution Partial Differential Equations Using Deep Learning*

A Comparative Study of the Explicit Finite Difference Method and Physics-Informed Neural Networks for Solving the Burgers' Equation

Multi-Grade Deep Learning for Partial Differential Equations with Applications to the Burgers Equation

Multi-Net Strategy: Accelerating Physics-Informed Neural Networks for Solving Partial Differential Equations.

Physics-Informed Neural Networks for Multi-Phase Flow in Porous Media Considering Dual Shocks and Interphase Solubility