WITHDRAWN Physics-Guided Unsupervised Deep Learning Approach for the Inversion of Receiver Functions in Dipping and Anisotropic Media
Description:
WITHDRAWN The converted wave technique namely Receiver function (RF) has been utilized routinely for probing the crust and mantle structures. In the analysis of receiver functions, deterministic physical methods such as inversions are frequently employed to refine the estimations to image the subsurface realistic geological structure. Despite their utility, the presence of dipping and anisotropic geological structures very often complicates and can even hinder the inversion process. To address these complexities, here we introduce a Physics-Guided unsupervised deep learning approach for the inversion of receiver function data. We employ unsupervised deep learning, enhanced with implicit neural representations, allowing for the prediction of inverted Earth model parameters: thickness (H), S-wave velocity (Vs), anisotropy, trend, plunge, strike and dip without requiring any labeled data. For determining the optimal model parameters, the output parameters are used in a forward modeling scheme to simulate the receiver functions. During training, the model iteratively adjusts and improves these parameters based on discrepancies between the simulated and observed receiver functions. Inversion results from both synthetic and field examples from the Indian shield suggest that physics-guided unsupervised deep learning approach is effective in inversion tasks, particularly when dealing with intricate geological settings like dipping and anisotropic media. With its application aimed at understanding subsurface structures, we believe that this approach holds potential to broaden the capabilities of subsurface exploration.
Session: Machine Learning for Full Waveform Inversion: From Hybrid to End-to-End Approaches [Poster Session]
Type: Poster
Date: 5/3/2024
Presentation Time: 08:00 AM (local time)
Presenting Author: Bijayananda
Student Presenter: Yes
Invited Presentation:
Authors
Bijayananda Dalai Presenting Author bijayanandadalai@gmail.com National Geophysical Research Institute |
Prakash Kumar Corresponding Author prakashk@ngri.res.in National Geophysical Research Institute |
Mrinal Sen mrinal@utexas.edu University of Texas at Austin |
|
|
|
|
|
|
WITHDRAWN Physics-Guided Unsupervised Deep Learning Approach for the Inversion of Receiver Functions in Dipping and Anisotropic Media
Category
Machine Learning for Full Waveform Inversion: From Hybrid to End-to-End Approaches