PINNup: Robust Neural Network Wavefield Solutions Using Frequency Upscaling and Neuron Splitting

Article Engineering, Multidisciplinary

On the eigenvector bias of Fourier feature networks: From regression to solving multi-scale PDEs with physics-informed neural networks

Sifan Wang et al.

Summary: PINNs show promise in integrating physical models with observational data, but struggle with high-frequency or multi-scale features. Through NTK theory, this work investigates the limitations of PINNs and proposes novel architectures to address these challenges.

COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING (2021)

Add to Collection

Article Computer Science, Interdisciplinary Applications

PINNeik: Eikonal solution using physics-informed neural networks

Umair bin Waheed et al.

Summary: The proposed algorithm based on physics-informed neural networks offers a high traveltime accuracy for a wide range of applications in seismology. It leverages machine learning techniques like transfer learning and surrogate modeling to speed up computations for updated velocity models and source locations. The method's flexibility in incorporating various physical properties and improving convergence rate makes it an efficient forward modeling engine for seismological applications.

COMPUTERS & GEOSCIENCES (2021)

Add to Collection

Article Geochemistry & Geophysics

Solving the frequency-domain acoustic VTI wave equation using physics-informed neural networks

Chao Song et al.

Summary: Frequency-domain wavefield solutions for the anisotropic acoustic wave equation can help describe the anisotropic nature of the Earth, but inverting the impedance matrix for such solutions can be computationally expensive. To address this challenge, the study proposes using physics-informed neural networks (PINNs) to obtain wavefield solutions for an acoustic wave equation in transversely isotropic (TI) media with a vertical axis of symmetry (VTI). By utilizing automatic differentiation in PINNs, numerical dispersion artefacts can be avoided, providing more accurate solutions.

GEOPHYSICAL JOURNAL INTERNATIONAL (2021)

Add to Collection

Article Physics, Mathematical