Abstract:The numerical approximation of solutions to the compressible Euler and Navier-Stokes equations is a crucial but challenging task with relevance in various fields of science and engineering. Recently, methods from deep learning have been successfully employed for solving partial differential equations by incorporating the equations into a loss function that is minimized during the training of a neural network. This approach yields a so-called physics-informed neural network. It is not based upon classical discretizations, such as finite-volume or finite-element schemes, and can even address parametric problems in a straightforward manner. This has raised the question, whether physics-informed neural networks may be a viable alternative to conventional methods for computational fluid dynamics. In this article we introduce an adaptive artificial viscosity reduction procedure for physics-informed neural networks enabling approximate parametric solutions for forward problems governed by the stationary two-dimensional Euler equations in sub- and supersonic conditions. To the best of our knowledge, this is the first time that the concept of artificial viscosity in physics-informed neural networks is successfully applied to a complex system of conservation laws in more than one dimension. Moreover, we highlight the unique ability of this method to solve forward problems in a continuous parameter space. The presented methodology takes the next step of bringing physics-informed neural networks closer towards realistic compressible flow applications.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to use Physics - Informed Neural Networks (PINNs) to solve the two - dimensional steady - state compressible Euler equations with parameterization. Specifically, the paper proposes an adaptive artificial viscosity reduction procedure to enable PINNs to approximately solve the parametric solutions in the forward problem, especially under subsonic and supersonic conditions. This is the first successful application of the concept of artificial viscosity to complex problems in multi - dimensional conservation law systems. In addition, the paper emphasizes the unique ability of this method to solve forward problems in the continuous parameter space, which is an important step towards applying PINNs to more realistic compressible fluid dynamics problems. ### Background and Objectives of the Paper 1. **Challenges in Numerical Solution**: The motion of compressible fluids leads to nonlinear partial differential equations (such as the Euler equations and the Navier - Stokes equations). The numerical solution of these equations is very important but extremely challenging in the fields of science and engineering. Although traditional numerical methods (such as the finite - volume method and the finite - element method) are already very mature, they have limitations in grid generation, computational efficiency, and hardware adaptation. 2. **Advantages of PINNs**: In recent years, deep - learning methods have been successfully used to solve partial differential equations by embedding the equations into the loss function and minimizing this loss function when training the neural network. This method is called Physics - Informed Neural Networks (PINNs). PINNs do not rely on traditional spatial or temporal discretization and can directly handle parameterized problems, that is, one network can be trained within a continuous parameter space and provide approximate solutions for a series of parameter combinations of interest. 3. **Research Objectives**: This paper aims to explore the application potential of PINNs in solving compressible fluid dynamics problems, especially by introducing adaptive artificial viscosity to improve the prediction accuracy of supersonic problems. The author hopes that through this method, a new approach can be provided for solving more complex forward problems. ### Main Contributions 1. **First Application**: For the first time, the concept of artificial viscosity has been successfully applied to complex problems in multi - dimensional conservation law systems, especially the two - dimensional steady - state compressible Euler equations. 2. **Adaptive Artificial Viscosity**: An adaptive artificial viscosity reduction procedure is proposed. By allowing the network to predict the locally required viscosity value and gradually reducing the viscosity during the training process, the error can be reduced while ensuring convergence. 3. **Parameterized Problems**: The ability of PINNs to solve forward problems in the continuous parameter space is demonstrated, which is of great significance for tasks such as design optimization. ### Method Overview 1. **Basic Principles of PINNs**: PINNs use neural networks as parametric approximation functions for the solutions of partial differential equations and optimize network parameters by minimizing the loss function that contains differential equations, initial conditions, and boundary conditions. 2. **Introduction of Artificial Viscosity**: An artificial viscosity term is introduced into the Euler equations in the form of: \[ \frac{\partial W}{\partial t}+\frac{\partial F_{x}}{\partial x}+\frac{\partial F_{y}}{\partial y}=\eta\Delta W \] where \(\Delta\) is the Laplace operator and \(\eta\) can be interpreted as artificial viscosity. The magnitude of the artificial viscosity is linearly scaled according to the wave speed \(a + |q|\): \[ \eta=\nu(a + |q|) \] where \(\nu\) is the viscosity factor. 3. **Adaptive Training Process**: A three - stage training process is proposed: - **First Stage**: Train using the ADAM optimizer at a fixed viscosity \(\tilde{\nu}=\nu_{0}\) until the residual loss no longer changes significantly. - **Second Stage**: Continue to use the ADAM optimizer to gradually reduce the predetermined viscosity \(\tilde{\nu}\) until \(\tilde{\nu} = 0\), and then continue training at \(\tilde{\nu}=0\) until the residual loss no longer changes significantly. - **Third Stage**: Use L

Physics-Informed Neural Networks for Parametric Compressible Euler Equations

Physics-Informed Neural Networks for Transonic Flows around an Airfoil

FEM-based Neural Networks for Solving Incompressible Fluid Flows and Related Inverse Problems

Physics-informed neural networks for high-speed flows

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

Inferring incompressible two-phase flow fields from the interface motion using physics-informed neural networks

Enhancing physics informed neural networks for solving Navier–Stokes equations

Physics-informed PointNet: A deep learning solver for steady-state incompressible flows and thermal fields on multiple sets of irregular geometries

Numerical solution of the system of Navier-Stokes equations in the case of a compressible medium using neural networks

Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations

Physics-driven Learning of the Steady Navier-Stokes Equations using Deep Convolutional Neural Networks

Active training of physics-informed neural networks to aggregate and interpolate parametric solutions to the Navier-Stokes equations

On Physics-Informed Deep Learning for Solving Navier-Stokes Equations

Physics-Informed Neural Networks for Solving Coupled Stokes-Darcy Equation

Physics-informed Autoencoders for Lyapunov-stable Fluid Flow Prediction

Learning the solution operator of two-dimensional incompressible Navier-Stokes equations using physics-aware convolutional neural networks

An artificial viscosity augmented physics-informed neural network for incompressible flow

Self-adaptive loss balanced Physics-informed neural networks for the incompressible Navier-Stokes equations

Can Physics-Informed Neural Networks beat the Finite Element Method?