Deep Neural Helmholtz Operators for 3D Elastic Wave Propagation and Inversion

Abstract

Numerical simulations of seismic wave propagation in heterogeneous 3D media are central to investigating subsurface structures and understanding earthquake processes, yet are computationally expensive for large problems. This is particularly problematic for full waveform inversion, which typically involves numerous runs of the forward process. In machine learning there has been considerable recent work in the area of operator learning, with a new class of models called neural operators allowing for data-driven solutions to partial differential equations. Recent works in seismology have shown that when neural operators are adequately trained, they can significantly shorten the compute time for wave propagation. However, the memory required for the 3D time domain equations may be prohibitive. In this study, we show that these limitations can be overcome by solving the wave equations in the frequency domain, also known as the Helmholtz equations, since the solutions for a set of frequencies can be determined in parallel. The 3D Helmholtz neural operator is 40 times more memory-efficient than an equivalent time-domain version. We employ a U-shaped neural operator for 2D and 3D elastic wave modeling, achieving two orders of magnitude acceleration compared to a baseline spectral element method. The neural operator accurately generalizes to variable velocity structures and can be evaluated on denser input meshes than used in the training simulations. We also show that when solving for wavefields strictly on the surface, the accuracy can be significantly improved via a graph neural operator layer. In leveraging automatic differentiation, the proposed method can serve as an alternative to the adjoint-state approach for 3D full-waveform inversion, reducing the computation time by a factor of 350.

References in corpus (11)

• DeepONet: Learning nonlinear operators for identifying differential equations based on the universal approximation theorem of operators
• Bayesian Deep Convolutional Encoder-Decoder Networks for Surrogate Modeling and Uncertainty Quantification
• Physics-informed Neural Networks (PINNs) for Wave Propagation and Full Waveform Inversions
• Solving the acoustic VTI wave equation using physics-informed neural networks
• Solving parametric PDE problems with artificial neural networks
• Fast approximate simulation of seismic waves with deep learning
• A General Approach to Seismic Inversion with Automatic Differentiation
• HypoSVI: Hypocenter inversion with Stein variational inference and Physics Informed Neural Networks
• PINNup: Robust neural network wavefield solutions using frequency upscaling and neuron splitting
• Fourier Neural Operator Surrogate Model to Predict 3D Seismic Waves Propagation
• Laplace HypoPINN: Physics-Informed Neural Network for hypocenter localization and its predictive uncertainty

Cited by in corpus (6)

• Multiple-Input Fourier Neural Operator (MIFNO) for source-dependent 3D elastodynamics
• Reduced-order modeling for complex 3D seismic wave propagation
• Ambient Noise Full Waveform Inversion with Neural Operators
• Forward and Inverse Mantle Convection with Neural Operators
• Waveform-Based Probabilistic Seismic Hazard Analysis Using Ground-Motion Generative Models
• Real-Time Reconstruction of Ground Motion During Small Magnitude Earthquakes: A Pilot Study