cs.LGMar 16, 2026

Seismic full-waveform inversion based on a physics-driven generative adversarial network

Authors: Xinyi Zhang, Caiyun Liu, Jie Xiong, Qingfeng Yu

Organizations: School of Electronic Information and Electrical Engineering, Yangtze University Jingzhou, Hubei 434023, P.R.China

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Objectives: Full-waveform inversion (FWI) is a high-resolution geophysical imaging technique that reconstructs subsurface velocity models by iteratively minimizing the misfit between predicted and observed seismic data. However, under complex geological conditions, conventional FWI suffers from strong dependence on the initial model and tends to produce unstable results when the data are sparse or contaminated by noise. Methods: To address these limitations, this paper proposes a physics-driven generative adversarial network-based full-waveform inversion method. The proposed approach integrates the data-driven capability of deep neural networks with the physical constraints imposed by the seismic wave equation, and employs adversarial training through a discriminator to enhance the stability and robustness of the inversion results. Results: Experimental results on two representative benchmark geological models demonstrate that the proposed method can effectively recover complex velocity structures and achieves superior performance in terms of structural similarity (SSIM) and signal-to-noise ratio (SNR). Conclusions: This method provides a promising solution for alleviating the initial-model dependence in full-waveform inversion and shows strong potential for practical applications.

Figures & tables

Figure 1: Overall workflow
Figure 2: Architecture of the physics-driven generator
Figure 3: Architecture of the discriminator: (a) Discriminator structure; (b) Convolutional block structure.
Figure 4: Inversion results of the Marmousi model: (a) True velocity model; (b) Gaussian-smoothed initial velocity model; (c) Inversion result by Yang et al. using noise-free seismic data; (d) Inversion result by the proposed method using noise-free seismic data; (e) Inversion result by Yang et al. using noisy seismic data; (f) Inversion result by the proposed method using noisy seismic data.
Noise conditionMetricProposed methodYang et al.’s method
Noise-freeSSIM0.70920.6802
 SNR19.3717.29
NoisySSIM0.66520.6301
 SNR21.4216.71
Table 1: Statistical performance of Marmousi model inversion
Figure 5: Inversion results of the Overthrust model: (a) True velocity model; (b) Gaussian-smoothed initial velocity model; (c) Inversion result by Yang et al. using noise-free seismic data; (d) Inversion result by the proposed method using noise-free seismic data; (e) Inversion result by Yang et al. using noisy seismic data; (f) Inversion result by the proposed method using noisy seismic data.
Noise conditionMetricProposed methodYang et al.’s method
Noise-freeSSIM0.78420.7270
 SNR28.5725.83
NoisySSIM0.72860.6889
 SNR27.5223.88
Table 2: Statistical performance of Overthrust model inversion
Figure 6: Inversion results using a linear initial model: (a1–d1) True velocity model, linear initial velocity model, inversion result by Yang et al., and inversion result by the proposed method for the Marmousi model; (a2–d2) True velocity model, linear initial velocity model, inversion result by Yang et al., and inversion result by the proposed method for the Overthrust model.
ModelMetricProposed methodYang et al.’s method
MarmousiSSIM0.65820.6550
 SNR18.2115.35
OverthrustSSIM0.76070.7377
 SNR28.5023.92
Table 3: Statistical performance of inversion using a linear initial model
More in this topic

Seismic Imaging

See all 15 papers

Explore similar work

CardsList

  1. Improving Full Waveform Inversion in Large Model Era

    Feb 27, 2026Yinan Feng, Peng Jin, Yuzhe Guo +2Inverse ProblemsOOD Generalization

    School of Data Science and Society University of North Carolina at Chapel Hill Chapel Hill, NC, USA · The Pennsylvania State University State College, PA, USA · Google DeepMind

  2. Full waveform inversion method based on diffusion model

    Mar 18, 2026Caiyun Liu, Siyang Pei, Qingfeng Yu +1Inverse ProblemsDiffusion-Based Inverse Problems

    School of Information and Mathematics, Yangtze University, Jingzhou, Hubei Province 434023, China · School of Electronic Information and Electrical Engineering, Yangtze University, Jingzhou, Hubei Province 434023, China

  3. Conditional Rectified Flow-based End-to-End Rapid Seismic Inversion Method

    Mar 16, 2026Haofei Xu, Wei Cheng, Sizhe Li +1Inverse ProblemsConditional Generative Modeling

    School of Electronic lnformation and Electrical Engineering, Yangtze UniversityJingzhou, Hubei 434023, PR.China