Abstract:Abstract Physics-informed neural networks (PINNs) have been used by researchers to solve partial differential equation (PDE)-constrained problems. We evaluate PINNs to solve for frequency-domain acoustic wavefields. PINNs can solely use PDEs to define the loss function for optimization without the need for labels. Partial derivatives of PDEs are calculated by mesh-free automatic differentiations. Thus, PINNs are free of numerical dispersion artifacts. It has been applied to the scattered acoustic wave equation, which relied on boundary conditions (BCs) provided by the background analytical wavefield. For a more direct implementation, we solve the nonscattered acoustic wave equation, avoiding limitations related to relying on the background homogeneous medium for BCs. Experiments support our following insights. Although solving time-domain wave equations using PINNs does not require absorbing boundary conditions (ABCs), ABCs are required to ensure a unique solution for PINNs that solve frequency-domain wave equations, because the single-frequency wavefield is not localized and contains wavefield information over the full domain. However, it is not trivial to include the ABC in the PINN implementation, so we develop an adaptive amplitude-scaled and phase-shifted sine activation function, which performs better than the previous implementations. Because there are only two outputs for the fully connected neural network (FCNN), we validate a linearly shrinking FCNN that can achieve a comparable and even better accuracy with a cheaper computational cost. However, there is a spectral bias problem, that is, PINNs learn low-frequency wavefields far more easily than higher frequencies, and the accuracy of higher frequency wavefields is often poor. Because the shapes of multifrequency wavefields are similar, we initialize the FCNN for higher frequency wavefields by that of the lower frequencies, partly mitigating the spectral bias problem. We further incorporate multiscale positional encoding to alleviate the spectral bias problem. We share our codes, data, and results via a public repository.

Physics-enhanced neural networks for half-space seismic wave modeling

Physics-constrained Neural Networks for Half-Space Seismic Wave Modeling

SeismicNet: Physics-informed neural networks for seismic wave modeling in semi-infinite domain

Physics-informed Neural Networks with Fourier Features for Seismic Wavefield Simulation in Time-Domain Nonsmooth Complex Media

Research progress of physics-informed neural network in seismic wave modeling

Physics-informed neural network for seismic wave inversion in layered semi-infinite domain

Self-adaptive physics-driven deep learning for seismic wave modeling in complex topography

Physics-informed Neural Networks (PINNs) for Wave Propagation and Full Waveform Inversions

Wave simulation in non-smooth media by PINN with quadratic neural network and PML condition

Accelerating the Convergence of Physics Informed Neural Networks for Seismic Wave Simulation

A Spatial-Temporally Adaptive PINN Framework for 3D Bi-Ventricular Electrophysiological Simulations and Parameter Inference

Numerical Solver-Independent Seismic Wave Simulation Using Task-Decomposed Physics-Informed Neural Networks.

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

Seismic Wavefields Modeling With Variable Horizontally Layered Velocity Models via Velocity-Encoded PINN

Physics-informed Neural Networks for Fully Non-Linear Free Surface Wave Propagation

Adaptive-Sampling Physics-Informed Neural Network for Viscoacoustic Wavefield Simulation

An Efficient Approximation Method Based on Enhanced Physics-Informed Neural Networks for Solving Localized Wave Solutions of PDEs

Physics-informed Neural Networks (PINN) for computational solid mechanics: Numerical frameworks and applications

Multiple Wavefield Solutions in Physics-Informed Neural Networks Using Latent Representation

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

Modeling multisource multifrequency acoustic wavefields by a multiscale Fourier feature physics-informed neural network with adaptive activation functions