Abstract:Full waveform inversion (FWI) reconstructs the underground velocity structures by minimizing the data residual between calculated wavefields and observed wavefileds. The conventional FWI usually uses some local optimization algorithms which lead to a strong dependency on initial velocity model. The objective function corresponding to low-frequency data components has less local minima. Reconstructing low-frequency information from recorded seismic data and using it in FWI can reduce cycle-skipping and thus weaken the dependency of inversion process on initial model. In this paper, based on the conventional frequency down-shifting method, we propose a sparse blind deconvolution-convolution low-frequency data reconstruction method, which can simultaneously update the wavelet and reconstruct the low-frequency components. First, we extract the subsurface reflection impulse responses (SRIR) by solving a Ll norm sparse constraint problem using the Fast Iterative Shrinkage-Thresholding Algorithm (FISTA). Then we test the accuracy of our algorithm, and discuss the effect of wavelet error and noises on the reconstruction result.When the wavelet is inaccurate, we update the amplitude of wavelet by alternately inverting Ll norm constraint and Tikhonov regularization problems, and correct the time-shift error by cross-correlating the direct waves. After that we can get the accurate wavelet and SRIR simultaneously. Then using the reconstructed data successively as observed data, combining it with dynamic random sources and layer-stripping methods, we propose a new strategy for the fast multiscale FWI. We test our method by numerical examples in several cases including blended acquisition cases. The results show that it has good anti-noise property and it can reconstruct valid low frequency components when the observed data lacking low-frequency information. The example using inaccurate wavelet shows that the blind deconvolution-convolution algorithm is able to obtain accurate wavelet and low-frequency data simultaneously. (C) 2017 Elsevier B.V. All rights reserved.

Layered low-frequency extrapolation with deep learning in full-waveform inversion

Extrapolated full-waveform inversion with deep learning

Deep Learning-Based Low-Frequency Extrapolation and Impedance Inversion of Seismic Data

Data-driven low-frequency signal recovery using deep-learning predictions in full-waveform inversion

Learning with real data without real labels: A strategy for extrapolated full-waveform inversion with field data

Efficient progressive transfer learning for full waveform inversion with extrapolated low-frequency reflection seismic data

A self-supervised learning framework for seismic low-frequency extrapolation

Transformer For Low-frequency Extrapolating of Seismic Data

Learning Seismic Low-Frequency Extrapolation in a Latent Space Using Limited Data

Prediction of Low-Frequency Seismic Data Without Relying on Wavelet Using Deep Learning

Low-Frequency Extrapolation of Prestack Viscoacoustic Seismic Data Based on Dense Convolutional Network

Full waveform inversion with extrapolated low frequency data

Low-Frequency Prediction Based on Multiscale and Cross-Scale Deep Networks in Full-Waveform Inversion.

Deep-learning assisted regularized elastic full waveform inversion using the velocity distribution information from wells

High-Resolution Regularized Elastic Full Waveform Inversion Assisted by Deep Learning

Deep-learning seismic full-waveform inversion for realistic structural models

Multiscale Data-driven Seismic Full-waveform Inversion with Field Data Study

Deep-Learning Full-Waveform Inversion Using Seismic Migration Images

Full-waveform inversion with adversarial losses via deep learning

An Augmented Lagrangian Method-Based Deep Iterative Unrolling Network for Seismic Full-Waveform Inversion

Sparse Blind Deconvolution Based Low-Frequency Seismic Data Reconstruction for Multiscale Full Waveform Inversion