Abstract:Neural networks have emerged as a tool for solving differential equations in many branches of engineering and science. But their progress in frequency domain acoustics is limited by the vanishing gradient problem that occurs at higher frequencies. This paper discusses a formulation that can address this issue. The problem of solving the governing differential equation along with the boundary conditions is posed as an unconstrained optimization problem. The acoustic field is approximated to the output of a neural network which is constructed in such a way that it always satisfies the boundary conditions. The applicability of the formulation is demonstrated on popular problems in plane wave acoustic theory. The predicted solution from the neural network formulation is compared with those obtained from the analytical solution. A good agreement is observed between the two solutions. The method of transfer learning to calculate the particle velocity from the existing acoustic pressure field is demonstrated with and without mean flow effects. The sensitivity of the training process to the choice of the activation function and the number of collocation points is studied.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is the vanishing gradient problem encountered when using neural networks to solve the Helmholtz equation in frequency - domain acoustics, especially in high - frequency cases. Specifically: 1. **Problems with existing methods**: The traditional Lagrange multiplier method will experience the vanishing gradient phenomenon at high frequencies when solving frequency - domain acoustic problems, causing the training process to be biased towards a certain loss function, thus affecting the prediction accuracy. 2. **Proposed method**: The author proposes a method based on physics - informed neural networks (PINNs) to solve these problems. By approximating the sound field as the output of a neural network and ensuring that it always satisfies the boundary conditions, the problem of solving partial differential equations and their boundary conditions is transformed into an unconstrained optimization problem. 3. **Scope of application**: This method is applicable to pipe acoustic problems with different cross - sectional shapes, including pipes with uniform cross - sections, pipes with gradually changing cross - sections, and thin tubes considering viscous - thermal effects. These application scenarios demonstrate the wide applicability and flexibility of this method. 4. **Specific objectives**: - Predict the sound field in pipes with uniform cross - sections. - Predict the sound field in pipes with gradually changing cross - sections. - Predict the complex - valued sound pressure in thin tubes (considering viscous - thermal effects). - Use transfer learning techniques to predict other acoustic variables (such as particle velocity) from the existing sound pressure fields. 5. **Innovative points**: - Adopt the trial solution approach to avoid the vanishing gradient problem at high frequencies. - Propose a method for handling the optimization problem of complex - valued sound fields by separating the real part and the imaginary part for optimization. - Verify the effectiveness of the neural network method by comparing with the analytical solution. In conclusion, this paper aims to provide a reliable and efficient neural - network - based method for solving complex problems in frequency - domain acoustics, especially for overcoming the vanishing gradient problem at high frequencies, thereby improving the prediction accuracy and stability.

Neural network based approach for solving problems in plane wave duct acoustics

Hard-constrained neural networks for modelling nonlinear acoustics

Practical Aspects of Physics-Informed Neural Networks Applied to Solve Frequency-Domain Acoustic Wave Forward Problem

Sound wave neural network based on partial differential equation

A Neural Network Warm-Start Approach for the Inverse Acoustic Obstacle Scattering Problem

Deep neural networks for waves assisted by the Wiener-Hopf method

Approximation of acoustic black holes with finite element mixed formulations and artificial neural network correction terms

Wavefield solutions from machine learned functions

Mathematics-Informed Neural Network for Acoustic Scattering by Parallel Plates

Learning based numerical methods for acoustic frequency-domain simulation with high frequency

Solving Inverse Problems in Steady State Navier-Stokes Equations using Deep Neural Networks

Artificial Neural Networks in Fluid Dynamics: A Novel Approach to the Navier-Stokes Equations

Physics-Informed Neural Networks for Transonic Flows around an Airfoil

Unsupervised neural networks for Maxwell fluid flow and heat transfer over a curved surface with nonlinear convection and temperature‐dependent properties

Galerkin Neural Network-POD for Acoustic and Electromagnetic Wave Propagation in Parametric Domains

A feedforward neural network framework for approximating the solutions to nonlinear ordinary differential equations

Helmholtz-equation Solution in Nonsmooth Media by a Physics-Informed Neural Network Incorporating Quadratic Terms and a Perfectly Matching Layer Condition

A feedforward neural network for modelling of average pressure frequency response

NeuralSound

Neural Network with Local Converging Input (NNLCI) for Supersonic Flow Problems with Unstructured Grids

Physics and geometry informed neural operator network with application to acoustic scattering