Full-Waveform Inversion With Velocity Model Low-Rank Implicit Neural Representation

Full-waveform inversion (FWI) is pivotal for exploring subsurface structures and physical parameters. However, classical FWI methods often experience challenges like cycle skipping and nonlinearity, necessitating accurate initial velocity models. Pure data-driven approaches based on deep learning ar...

Full description

Saved in:
Bibliographic Details
Published in:IEEE transactions on geoscience and remote sensing Vol. 63; pp. 1 - 16
Main Authors: Chen, Ruihua, Wu, Bangyu, Li, Meng, Luo, Yisi
Format: Journal Article
Language:English
Published: New York IEEE 2025
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Subjects:
Accuracy
Convolutional neural networks
Data models
Deep learning
Full-waveform inversion (FWI)
implicit neural representation (INR)
Machine learning
matrix factorization
Neural coding
Neural networks
Noise
Nonlinear systems
Nonlinearity
Numerical models
Physical properties
Physics
physics-guided velocity model inversion
Predictive models
Representations
Robustness
Seismic data
Smoothness
Sun
Training
Training data
Velocity
velocity model reparameterization
Waveforms
ISSN:0196-2892, 1558-0644
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Full-waveform inversion (FWI) is pivotal for exploring subsurface structures and physical parameters. However, classical FWI methods often experience challenges like cycle skipping and nonlinearity, necessitating accurate initial velocity models. Pure data-driven approaches based on deep learning are constrained by limited real labeled data and inadequate generalization, hindering practical applications. To address these issues, we propose an unsupervised physics-informed machine learning (PIML) FWI with low-rank implicit neural representation (INR, termed LR-IFWI), which utilizes a low-rank matrix factorization parameterized by coordinate-based neural networks, continuously and compactly representing the velocity model by implicitly encoding low-rank properties and smoothness. Our method has three crucial advantages: 1) LR-IFWI considerably improves inversion accuracy and shows steadier inversion convergence behavior; 2) LR-IFWI can reduce the number of iterations for comparable inversion accuracy, improving efficiency attributed to the compact low-rank representation; and 3) LR-IFWI has better robustness, alleviating the dependence on the initial models and improving noise resistance due to the physical constraints and low-rank neural representation. Numerical tests on the 2-D Marmousi model demonstrate that LR-IFWI achieves efficient and accurate inversion with fewer iterations and greater precision when utilizing smooth, linear, and random initial velocity models. Further experiments with noisy seismic data, missing low-frequency components, and more challenging velocity models, such as 2-D SEG/EAGE salt and overthrust models, highlight its robustness and generalization.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0196-2892
1558-0644
DOI:10.1109/TGRS.2025.3594184