Progressive transfer learning for low-frequency data prediction in full-waveform inversion
نویسندگان
چکیده
To effectively overcome the cycle-skipping issue in full-waveform inversion (FWI), we have developed a deep neural network (DNN) approach to predict absent low-frequency (LF) components by exploiting hidden physical relation connecting LF and high-frequency (HF) data. efficiently solve this challenging nonlinear regression problem, two novel strategies are proposed design DNN architecture optimize learning process: (1) dual data feed structure (2) progressive transfer learning. With structure, not only HF data, but also corresponding beat tone fed into relieve burden of feature extraction. The second strategy, learning, enables us train using single evolving training set. Within framework set continuously evolves an iterative manner gradually retrieving subsurface information through physics-based module, progressively enhancing prediction accuracy propelling process out local minima. synthetic numerical experiments suggest that, without any priori geologic information, predicted sufficiently accurate for FWI engine produce reliable velocity models free artifacts.
منابع مشابه
Full waveform inversion with extrapolated low frequency data
The availability of low frequency data is an important factor in the success of full waveform inversion (FWI) in the acoustic regime. The low frequencies help determine the kinematically relevant, low-wavenumber components of the velocity model, which are in turn needed to avoid convergence of FWI to spurious local minima. However, acquiring data below 2 or 3 Hz from the field is a challenging ...
متن کامل3D acoustic frequency-domain full-waveform inversion
We present one of the first attempt at implementing a massively parallel frequency-domain full-waveform inversion algorithm for imaging 3D acoustic media. The inverse method is based on a classic steepest-descent algorithm. The algorithm was designed so that one or several frequencies are inverted at a time. Wave propagation modeling, a key component of the inversion algorithm, is performed wit...
متن کاملDiscretized Adjoint State Time and Frequency Domain Full Waveform Inversion: A Comparative Study
This study derives the discretized adjoint states full waveform inversion (FWI) in both time and frequency domains based on the Lagrange multiplier method. To achieve this, we applied adjoint state inversion on the discretized wave equation in both time domain and frequency domain. Besides, in this article, we introduce reliability tests to show that the inversion is performing as it should be ...
متن کاملImage-guided full waveform inversion
Multiple problems, including high computational cost, spurious local minima, and solutions with no geologic sense, have prevented widespread application of full waveform inversion (FWI), especially FWI of seismic reflections. These problems are fundamentally related to a large number of model parameters and to the absence of low frequencies in recorded seismograms. Instead of inverting for all ...
متن کاملA projected Hessian for full waveform inversion
A Hessian matrix in full waveform inversion (FWI) is difficult to compute directly because of high computational cost and an especially large memory requirement. Therefore, Newton-like methods are rarely feasible in realistic largesize FWI problems. We modify the BFGS method to use a projected Hessian matrix that reduces both the computational cost and memory required, thereby making a quasiNew...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Geophysics
سال: 2021
ISSN: ['0016-8033', '1942-2156']
DOI: https://doi.org/10.1190/geo2020-0598.1