Article
Geochemistry & Geophysics
Wenqian Fang, Lihua Fu, Shaoyong Liu, Hongwei Li
Summary: Deep-learning technology has enabled automatic seismic data interpolation by learning the mapping between regularly subsampled and complete data. However, limitations such as inadequate generalization performance and difficulty in network interpretation exist. To address these issues, PEFNet combines deep neural networks and classic PEF methods, offering improved local dip extraction and good generalization ability.
Article
Geochemistry & Geophysics
Thomas Andre Larsen Greiner, Jan Erik Lie, Odd Kolbjornsen, Andreas Kjelsrud Evensen, Espen Harris Nilsen, Hao Zhao, Vasily Demyanov, Leiv-J. Gelius
Summary: In 3D marine seismic acquisition, the seismic wavefield is not uniformly sampled. This study proposes an unsupervised deep learning approach based on a convolutional neural network to reconstruct the incomplete seismic wavefield. The method utilizes appropriate regularization penalties and exhibits improved performance compared to traditional approaches in both synthetic and field data examples.
Article
Geochemistry & Geophysics
Feng Qian, Yan Wang, Bingwei Zheng, Zhangbo Liu, Yingjie Zhou, Guangmin Hu
Summary: In this study, a framelet-based order-p tensor neural network model is proposed for learning the priors of multidimensional seismic images. By redefining the framelet and the order-p t-product, we can effectively denoise higher dimensional seismic data in a data-driven manner. Experimental results demonstrate the advantages of our method on synthetic and real field seismic datasets compared to other state-of-the-art methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geosciences, Multidisciplinary
Yuri S. F. Bezerra, German Garabito, Mauricio D. Sacchi
Summary: This paper presents and compares various Fourier reconstruction algorithms for seismic data processing, discussing their strengths and weaknesses. Through experiments and tests, the effectiveness of Fourier techniques in dealing with sparse and low-quality seismic data is validated, with a particular focus on using Fourier interpolation for data preconditioning before prestack time migration.
JOURNAL OF APPLIED GEOPHYSICS
(2021)
Article
Geochemistry & Geophysics
Xiao Niu, Lihua Fu, Wenqian Fang, Qin Wang, Meng Zhang
Summary: The use of nonlocal self-similarity (NSS) prior has been effective in attenuating random noise in seismic data due to the repetitiveness of textures and structures. However, NSS-based approaches face challenges in seismic interpolation, especially when dealing with missing traces. To solve this problem, a two-stage iterative seismic-interpolation framework based on a rank-reduction (RR) algorithm is developed. Our approach can handle irregularly or regularly sampled seismic data and achieves enhanced interpolation performance compared with other methods.
Article
Computer Science, Information Systems
Emad B. Helal, Omar M. Saad, Ali G. Hafez, Yangkang Chen, Gamal M. Dousoky
Summary: This paper proposes two convolutional autoencoders for seismic data compression, based on convolutional neural networks, which efficiently utilize bandwidth allocation and reconstruct input seismic data perfectly. The first model is effective at low compression ratios, while the second model improves signal-to-noise ratio at moderate and high compression ratios compared to benchmark algorithms.
Article
Geochemistry & Geophysics
Xuemin Zhang, Jianwei Ma, Hao Zhang
Summary: Seismic data can be described in a low-dimensional manifold, which is a good regularizer for interpolation. However, low-dimensional manifold regularization alone is not sufficient for interpolating seismic data with large data gaps or spectral aliasing. Therefore, a curvature-regularized low-dimensional manifold method is proposed for seismic data interpolation. The proposed method outperforms other methods in many missing data cases, especially in the presence of large gaps or spectral aliasing, as shown by numerical experiments on synthetic and field data.
Article
Geochemistry & Geophysics
Saber Jahanjooy, Mohammad Ali Riahi, Hamed Ghanbarnejad Moghanloo
Summary: The acoustic impedance model is crucial in seismic interpretation, and commonly used regularization methods like Tikhonov and total variation have limitations in complex models. A sequential combination of Tikhonov and total variation regularization can provide more accurate results.
Article
Geochemistry & Geophysics
Jonathan Popa, Susan E. Minkoff, Yifei Lou
Summary: Seismic data are often incomplete due to various reasons, but missing data can be recovered using tensor completion methods. By utilizing tensor singular value decomposition and exploiting the conjugate symmetry of multidimensional Fourier transform, the effectiveness and efficiency of low-rank data reconstruction can be improved. Our improved tSVD algorithm, which incorporates conjugate symmetry, reduces the runtime of the inner loop and shows faster computation compared to traditional methods.
Article
Chemistry, Multidisciplinary
Jiyun Yu, Daeung Yoon
Summary: In seismic data acquisition, data loss can occur in marine seismic exploration due to the use of streamer systems. Machine learning techniques have been used to improve the resolution of seismic data in the crossline direction. This study introduces a 3D cWGAN for interpolating 3D seismic data and compares its performance with 2D cWGAN and 3D U-Net. The results show that the 3D cWGAN is more efficient in enhancing resolution and computation.
APPLIED SCIENCES-BASEL
(2023)
Article
Geochemistry & Geophysics
Pengfei Yu, Mingzhi Chu, Jiameng Jiang
Summary: Proper decomposition of P- and S-waves is crucial for multicomponent ocean bottom seismic data processing. Separate calibration filters are designed to accurately calibrate the kinematic and dynamic characteristics of particle velocity (or displacement) components, allowing for appropriate adjustment of energy distribution within each component.
PURE AND APPLIED GEOPHYSICS
(2023)
Article
Geochemistry & Geophysics
Jie Shao, Yibo Wang, Xu Chang
Summary: The study utilized a four-step Radon domain interferometric interpolation method to process sparse seismic data, which involved generating synthetic seismic data, transforming into the Radon domain, conducting seismic interferometric interpolation, and applying a matching filter. The method effectively interpolated sparse seismic data, yielding more accurate results compared to traditional methods and showing less sensitivity to water layer depth and impedance contrast of the water bottom.
Article
Geochemistry & Geophysics
Yangkang Chen, Sergey Fomel, Hang Wang, Shaohuan Zu
Summary: The article introduces the use of prediction-error filter (PEF) and nonstationary prediction-error filter (NPEF) in seismic data processing, solving highly ill-posed inverse problems through computationally efficient iterative shaping regularization. NPEF can be used for denoising and restoring irregularly missing and regularly missing 5D seismic data.
Article
Geosciences, Multidisciplinary
Federico Mori, Amerigo Mendicelli, Gaetano Falcone, Gianluca Acunzo, Rose Line Spacagna, Giuseppe Naso, Massimiliano Moscatelli
Summary: Past seismic events have shown that the damage and casualties caused by earthquakes depend on both the strength of the ground motion and the local site conditions. A machine learning approach was used to predict ground motion using geological and geophysical data, resulting in maps with high resolution and accuracy. These predictions are crucial for supporting rescue operations and interventions.
NATURAL HAZARDS AND EARTH SYSTEM SCIENCES
(2022)
Article
Thermodynamics
Chun Fu, Matias Quintana, Zoltan Nagy, Clayton Miller
Summary: Building energy prediction and management is increasingly important, and the inconsistency and incompleteness of energy data hinder accurate predictions and management. This study proposes using image-based deep learning method, specifically Partial Convolution (PConv), to impute missing energy data and demonstrates its effectiveness in generating more accurate predictions. The results show that PConv outperforms other methods and offers a scalable and effective solution for filling in missing energy data.
APPLIED THERMAL ENGINEERING
(2023)
Article
Geochemistry & Geophysics
Rongzhi Lin, Breno Bahia, Mauricio D. Sacchi
Summary: In this study, the simultaneous source separation problem is addressed using the projected gradient descent (PGD) method with a windowed robust singular spectrum analysis (SSA) filter to suppress source interferences. The SSA filter is reformulated as a robust optimization problem solved with a bifactored gradient descent (BFGD) algorithm, enhancing robustness by incorporating Tukey's biweight loss function. The proposed robust SSA filter shows decreased sensitivity to rank-selection and accelerates the convergence of the PGD method compared to classical SSA filters.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Kristian Torres, Mauricio Sacchi
Summary: This paper addresses two common issues in least-squares reverse time migration (LSRTM), namely the need for many iterations to improve subsurface imaging and the difficulty of choosing appropriate regularization strategies. By combining supervised learning and deep learning, the proposed method achieves accurate estimation of reflectivity distributions in a few iterations, overcoming these drawbacks. Transfer learning is also demonstrated on real field data, producing high-resolution images with reduced iteration requirements compared to traditional LSRTM.
Article
Geochemistry & Geophysics
Dawei Liu, Xiaokai Wang, Xiaohai Yang, Haibo Mao, Mauricio D. Sacchi, Wenchao Chen
Summary: In recent years, supervised deep learning has been widely used in seismic processing. However, the weak generalization behavior restricts its implementation on large-scale prestack data sets for noise attenuation. To address this issue, this paper combines deep learning with an offset-vector tile (OVT) partitioning method to suppress strong scattered noise. Experimental results demonstrate that this method has excellent generalization ability.
Article
Geochemistry & Geophysics
Dawei Liu, Wei Wang, Xiaokai Wang, Zhensheng Shi, Mauricio D. Sacchi, Wenchao Chen
Summary: Revealing hidden reservoirs shielded by strong background interference (SBI) is critical to refined interpretation. A workflow using sparse representation and deep learning is developed to suppress SBI and enhance reservoir structures. Field data experiments demonstrate the success of the workflow in removing SBI and improving reservoir characterization.
Article
Geochemistry & Geophysics
Rongzhi Lin, Yi Guo, Fernanda Carozzi, Mauricio D. Sacchi
Summary: This study proposes an iterative rank reduction algorithm implemented via multichannel singular spectrum analysis (MSSA) filtering for data deblending. The original algorithm is limited to operating on data deployed on a regular grid, and this study introduces a modification using interpolated-MSSA for data with arbitrary irregular-grid coordinates. The results demonstrate that the proposed algorithm can successfully deblend and reconstruct sources simultaneously.
Article
Geochemistry & Geophysics
Hongling Chen, Mauricio D. Sacchi, Hojjat Haghshenas Lari, Jinghuai Gao, Xiudi Jiang
Summary: Reflectivity inversion methods are essential for seismic data processing. Unfortunately, they do not apply to nonstationary deconvolution cases. Therefore, we adopt a semisupervised deep learning approach to invert reflectivity when the propagating wavelet is considered unknown and time-variant.
Editorial Material
Geochemistry & Geophysics
Jianwei Ma, Qiuming Cheng, Sergey Fomel, Mauricio Sacchi, Ru-Shan Wu, Yike Liu, Yunyue Elita Li
Article
Geochemistry & Geophysics
Hanh Bui, Mirko van der Baan, Mauricio D. Sacchi
Summary: This study investigates two time-frequency methods, namely the sparse Gabor transform and the neighboring block thresholding, for enhanced event detection and selection. The sparse Gabor transform attenuates noise significantly while preserving signals, while the neighboring block thresholding leads to amplitude fidelity issues in the reconstructed waveforms.
Article
Geochemistry & Geophysics
Fan Li, Jianjun Gao, Mauricio D. Sacchi, Jun Lu, Yun Wang, Haifeng Chen, Chaolin Li
Summary: Multicomponent seismic exploration is attracting attention in the search for and identification of subtle oil and gas reservoirs. However, the sparse and irregular sampling problems encountered in single-component acquisition also apply to multicomponent seismic data acquisition. In this study, we propose a new vector Projection Onto Convex Sets (POCS) method with biquaternion Fourier transform to reconstruct irregularly missing traces of 3-D and three-component (3-D-3C) seismic data. The proposed method performs in the frequency domain and maintains the conjugate symmetry property of the biquaternion data, demonstrating its effectiveness and superiority over existing methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Rafael Manenti, Mauricio D. Sacchi
Summary: Tensors are used to represent pre-stack seismic data, especially for denoising and reconstruction. The tensor tree decomposition methodology allows for decomposing high-order tensors into third-order tensors, which can effectively preserve information while reducing the space needed. An algorithm that utilizes the tensor tree for data reconstruction was developed and compared to parallel matrix factorization, and it performed just as well in reconstructing both synthetic and field data.
GEOPHYSICAL PROSPECTING
(2023)
Correction
Geochemistry & Geophysics
Fan Li, Jianjun Gao, Mauricio D. Sacchi, Jun Lu, Yun Wang, Haifeng Chen, Chaolin Li
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Yi Guo, Rongzhi Lin, Mauricio D. Sacchi
Summary: Seismic acquisition costs depend on the number of sensors used. Limiting sensors can be beneficial, especially when they are expensive to purchase, deploy, and maintain. This study proposes an optimal design method for ocean bottom node (OBN) detector deployment using reinforcement learning approach. It utilizes an initial dataset to extract a prelearned basis library, and then uses Q-learning to find the sensor configuration that maximizes reconstruction quality.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Hongling Chen, Mauricio D. Sacchi, Jinghuai Gao
Summary: Researchers propose a new convolutional dictionary learning algorithm by introducing a parametric constraint to simplify filters and achieve a more efficient and structured representation of data. Experimental results demonstrate that the proposed method achieves superior reconstruction results.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Daniel S. Brox, Mauricio D. Sacchi
Summary: This article introduces a robust vector multichannel singular spectrum analysis (MSSA) algorithm for denoising seismic data, and presents the initial results of applying this algorithm to synthetic and real seismic data records. The MSSA algorithm, originally used for denoising scalar seismic wavefields, is now generalized to denoise vector seismic wavefields. Additionally, a robust rank reduction algorithm is introduced within the MSSA denoising algorithm to attenuate erratic signals in the input wavefield without distorting the output.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Dawei Liu, Mauricio D. Sacchi, Wenchao Chen
Summary: The study on five-dimensional seismic reconstruction is gaining attention, and this paper proposes two efficient methods to utilize the low rank structure of tensors and reduce computational costs.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
