EUVIMEDEUVIMED entdecken
Crossref

A Frequency‐Progressive Wavelet Packet‐Style Misfit Function With CNN‐Based Velocity Representation Extension for Full‐Waveform Inversion

Yijun Xi, Yu Lei, Chao Yang, Jinxin Hao, Xiongfeng Ren, Yali Zeng, Gai Gao

Geophysical Prospecting · 2026-11 · 74 (9)

DOI: 10.1111/1365-2478.70284 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Geophysical Prospecting, online erschienen am 13.11.2026. Verfasst von Yijun Xi und 6 weiteren Autor:innen. Die Arbeit nennt 29 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Cycle skipping is one of the main challenges limiting the convergence and reliability of full‐waveform inversion (FWI), especially when the initial velocity model is insufficiently accurate. To address this issue, we propose a frequency‐progressive wavelet packet‐style misfit function that exploits the multiscale frequency representation of seismic data to guide the inversion progressively from low‐ to high‐frequency information. Specifically, both observed and predicted seismic data are recursively decomposed into multiple wavelet packet‐style components corresponding to different frequency bands. During the early stages of inversion, greater weights are assigned to low‐frequency components and smaller weights to high‐frequency components, enabling the recovery of large‐scale velocity structures and reducing the likelihood of convergence to local minima. As the inversion progresses, the weights are adaptively adjusted by gradually decreasing the contribution of low‐frequency components and increasing that of high‐frequency components until all components are equally weighted, thereby facilitating the reconstruction of fine‐scale model details. Furthermore, a convolutional neural network (CNN)‐based velocity representation extension is introduced as a regularization strategy to improve inversion robustness and accuracy. Numerical experiments on synthetic and field data demonstrate that the proposed method effectively mitigates cycle skipping and yields more accurate velocity models than conventional FWI.

Zitieren

APA

Xi, Y., Lei, Y., Yang, C., Hao, J., Ren, X., Zeng, Y., & Gao, G. (2026). A Frequency‐Progressive Wavelet Packet‐Style Misfit Function With CNN‐Based Velocity Representation Extension for Full‐Waveform Inversion. Geophysical Prospecting, 74(9). https://doi.org/10.1111/1365-2478.70284

Vancouver

Xi Y, Lei Y, Yang C, Hao J, Ren X, Zeng Y, et al. A Frequency‐Progressive Wavelet Packet‐Style Misfit Function With CNN‐Based Velocity Representation Extension for Full‐Waveform Inversion. Geophysical Prospecting. 2026;74(9). doi:10.1111/1365-2478.70284

BibTeX
@article{Xi2026,
  author = {Xi, Yijun and Lei, Yu and Yang, Chao and Hao, Jinxin and Ren, Xiongfeng and Zeng, Yali and Gao, Gai},
  title = {A Frequency‐Progressive Wavelet Packet‐Style Misfit Function With CNN‐Based Velocity Representation Extension for Full‐Waveform Inversion},
  journal = {Geophysical Prospecting},
  year = {2026},
  volume = {74},
  number = {9},
  doi = {10.1111/1365-2478.70284}
}
Kostenlos starten →