Abstract:Solving partial differential equations (PDEs) is essential in scientific forecasting and fluid dynamics. Traditional approaches often incur expensive computational costs and trade-offs in efficiency and accuracy. Recent deep neural networks improve accuracy but require quality training data. Physics-informed neural networks (PINNs) effectively integrate physical laws, reducing data reliance in limited sample scenarios. A novel machine-learning framework, Chebyshev physics-informed Kolmogorov-Arnold network (ChebPIKAN), is proposed to integrate the robust architectures of Kolmogorov-Arnold networks (KAN) with physical constraints to enhance calculation accuracy of PDEs for fluid mechanics. We explore the fundamentals of KAN, emphasis on the advantages of using the orthogonality of Chebyshev polynomial basis functions in spline fitting, and describe the incorporation of physics-informed loss functions tailored to specific PDEs in fluid dynamics, including Allen-Cahn equation, nonlinear Burgers equation, two-dimensional Helmholtz equations, two-dimensional Kovasznay flow and two-dimensional Navier-Stokes equations. Extensive experiments demonstrate that the proposed ChebPIKAN model significantly outperforms standard KAN architecture in solving various PDEs by embedding essential physical information more effectively. These results indicate that augmenting KAN with physical constraints can not only alleviate overfitting issues of KAN but also improve extrapolation performance. Consequently, this study highlights the potential of ChebPIKAN as a powerful tool in computational fluid dynamics, proposing a path toward fast and reliable predictions in fluid mechanics and beyond.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the computational efficiency and accuracy issues in solving partial differential equations (PDEs) in fluid mechanics. Traditional methods often require a large amount of computational resources when solving PDEs, and they sacrifice accuracy while improving computational efficiency. Although deep neural networks perform well in improving accuracy, they rely on high - quality training data. And the physics - informed neural networks (PINNs) reduce the dependence on a large amount of sample data by integrating physical laws, but still have problems such as vanishing gradients and hyper - parameter sensitivity in some cases. To overcome these challenges, the paper proposes a new machine - learning framework - the physics - informed Kolmogorov - Arnold network (ChebPIKAN) based on Chebyshev polynomial - based basis functions. This framework combines the robust architecture of the Kolmogorov - Arnold network (KAN) and physical constraints, aiming to enhance the computational accuracy of PDEs in fluid mechanics. Specifically, the paper explores the basic principles of KAN, emphasizes the advantages of using the orthogonality of Chebyshev polynomial basis functions in spline fitting, and describes how to design physics - informed loss functions for specific fluid - mechanics PDEs (such as the Allen - Cahn equation, the nonlinear Burgers equation, the two - dimensional Helmholtz equation, the two - dimensional Kovasznay flow, and the two - dimensional Navier - Stokes equation). Through extensive experimental verification, the paper shows that the proposed ChebPIKAN model significantly outperforms the standard KAN architecture in solving various PDEs and can more effectively embed crucial physical information. This not only alleviates the over - fitting problem of KAN but also improves the extrapolation performance. Therefore, this research highlights the potential of ChebPIKAN in computational fluid mechanics and provides a new path for fast and reliable flow - field prediction.

Physics-informed Kolmogorov-Arnold Network with Chebyshev Polynomials for Fluid Mechanics

Kolmogorov Arnold Informed neural network: A physics-informed deep learning framework for solving PDEs based on Kolmogorov Arnold Networks

Physics Informed Kolmogorov-Arnold Neural Networks for Dynamical Analysis via Efficent-KAN and WAV-KAN

Kolmogorov Arnold Informed neural network: A physics-informed deep learning framework for solving forward and inverse problems based on Kolmogorov Arnold Networks

SPIKANs: Separable Physics-Informed Kolmogorov-Arnold Networks

Adaptive Training of Grid-Dependent Physics-Informed Kolmogorov-Arnold Networks

Physics-Informed Kolmogorov-Arnold Networks for Power System Dynamics

Physics-Based Deep Learning for Flow Problems

Deep learning based on mixed-variable physics informed neural network for solving fluid dynamics without simulation data

Applying physics informed neural network for flow data assimilation

Advancing fluid dynamics simulations: A comprehensive approach to optimizing physics-informed neural networks

Physically Plausible and Conservative Solutions to Navier-Stokes Equations Using Physics-Informed CNNs

From PINNs to PIKANs: Recent Advances in Physics-Informed Machine Learning

A High-Efficient Hybrid Physics-Informed Neural Networks Based on Convolutional Neural Network

Physics-Informed Deep Learning For Incompressible Laminar Flows

Physics-informed neural networks (PINNs) for fluid mechanics: A review

Frequency-compensated PINNs for Fluid-dynamic Design Problems

A comprehensive and FAIR comparison between MLP and KAN representations for differential equations and operator networks

Physics-informed neural networks for solving forward and inverse flow problems via the Boltzmann-BGK formulation

Combining physics-informed graph neural network and finite difference for solving forward and inverse spatiotemporal PDEs

Application of Physics-Informed Neural Network on the Example of Modeling Hydrodynamic Processes that Allow an Analytical Solution