Deep learning-driven velocity model building workflow

Araya‐Polo, Mauricio; Farris, Stuart; Florez, M. A.

doi:10.1190/tle38110872a1.1

Cited by 59 publications

(18 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Their data assimilation-driven loop at its core has a CNN-based network to compute prior models. It is conceptually comparable to FWI since their approach transforms data into models through an iterative process driven by a misfit where gradients are calculated using priors predicted by the CNN, therefore it is also related to works such as [48] and [39]. The network architecture is composed by 2 2D CNN and 4 FC layers, dropout is used and the activation function through the architecture is tanh.…”

Section: Semblance Velocity Analysismentioning

confidence: 99%

“…In Araya-Polo et al [39] a 2D CNN-based GAN architecture is used to generate velocity models that then through modeling generate seismic data which in turn become the training data for a FCN that reconstructs velocity models. This FCN through transfer learning increases accuracy as this generation-physics-prediction workflow iterates.…”

Section: B Deep Generative Modelingmentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning for Seismic Inverse Problems: Toward the Acceleration of Geophysical Analysis Workflows

Adler¹,

Araya‐Polo²,

Poggio³

2021

IEEE Signal Process. Mag.

Self Cite

View full text Add to dashboard Cite

Seismic inversion is a fundamental tool in geophysical analysis, providing a window into the Earth. In particular, it enables the reconstruction of large scale subsurface earth models for hydrocarbon exploration, mining, earthquakes analysis, shallow hazard assessment and other geophysical tasks. This article provides a comprehensive and timely overview of emerging datadriven Deep Learning (DL) solutions to seismic inverse problems, including velocity, impedance, reflectivity model building and seismic bandwidth extension. The article reviews seismic wave propagation and signal acquisition principles using large scale sensor arrays in offshore and inland exploration. In addition, the relations between the seismic forward and inverse problems are established, and prominent model-based solutions including seismic tomography, Full-Waveform Inversion (FWI) and multiscale FWI are explained. A primer to DL principles is presented, including empirical risk minimization, stochastic gradient-based learning, Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN) and regularization. The article then presents DL-based solutions addressing full seismic inversion as well as DL-assisted FWI solutions including Encoder-Decoder architectures, CNN-based networks and RNN-based networks with Long Short Term Memory and Gated Recurrent Units. Seismic signals features extraction methods such as semblance velocity analysis and seismic spectrograms are discussed, as well as batch vs. sequential data processing architectures. The article further presents advanced solutions based on Generative Adversarial Networks , and Physics-guided DL architectures that incorporate numerical geophysical modeling into the DL training loop, enabling unsupervised and semi-supervised learning.

show abstract

Section: Semblance Velocity Analysismentioning

confidence: 99%

Section: B Deep Generative Modelingmentioning

confidence: 99%

Deep Learning for Seismic Inverse Problems: Toward the Acceleration of Geophysical Analysis Workflows

Adler¹,

Araya‐Polo²,

Poggio³

2021

IEEE Signal Process. Mag.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Seismic exploration acquires data continuously in the horizontal direction; thus, it has the advantage of generating data with improved horizontal resolution relative to that obtained by conventional probebased oceanographic methods (Dagnino et al, 2016). Therefore, SO is used to image the structure of water layers (Tsuji et al, 2005;Sheen et al, 2012;Piété et al, 2013;Moon et al, 2017) and provide quantitative information, such as the physical properties (i.e., temperature, salinity) (Papenberg et al, 2010;Blacic et al, 2016;Dagnino et al, 2016;Jun et al, 2019) or the spectral distribution of the internal waves and turbulence (Sheen et al, 2009;Holbrook et al, 2013;Fortin et al, 2016) after analysis where temperature or salinity contrasts produce clear seismic reflections.…”

Section: Introductionmentioning

confidence: 99%

“…The use of artificial intelligence (AI) has been studied in geophysics for decades (McCormack, 1991;McCormack et al, 1993;Van der Baan and Jutten, 2000), but recent advances in computer resources and algorithms have spurred AI research, and several studies have been conducted to apply machine learning in the field of seismic data processing (Araya-Polo et al, 2019;Yang and Ma, 2019;Zhao et al, 2019). Among them, one of the most actively studied areas is pre-stack and post-stack data noise attenuation.…”

Section: Introductionmentioning

confidence: 99%

Random noise attenuation of sparker seismic oceanography data with machine learning

et al. 2020

View full text Add to dashboard Cite

Abstract. Seismic oceanography (SO) acquires water column reflections using controlled source seismology and provides high lateral resolution that enables the tracking of the thermohaline structure of the oceans. Most SO studies obtain data using air guns, which can produce acoustic energy below 100 Hz bandwidth, with vertical resolution of approximately 10 m or more. For higher-frequency bands, with vertical resolution ranging from several centimeters to several meters, a smaller, low-cost seismic exploration system may be used, such as a sparker source with central frequencies of 250 Hz or higher. However, the sparker source has a relatively low energy compared to air guns and consequently produces data with a lower signal-to-noise (S∕N) ratio. To attenuate the random noise and extract reliable signal from the low S∕N ratio of sparker SO data without distorting the true shape and amplitude of water column reflections, we applied machine learning. Specifically, we used a denoising convolutional neural network (DnCNN) that efficiently suppresses random noise in a natural image. One of the most important factors of machine learning is the generation of an appropriate training dataset. We generated two different training datasets using synthetic and field data. Models trained with the different training datasets were applied to the test data, and the denoised results were quantitatively compared. To demonstrate the technique, the trained models were applied to an SO sparker seismic dataset acquired in the Ulleung Basin, East Sea (Sea of Japan), and the denoised seismic sections were evaluated. The results show that machine learning can successfully attenuate the random noise in sparker water column seismic reflection data.

show abstract

“…Recently, rapid advances in computer resources have spurred artificial intelligence (AI) research, and several studies have been conducted to apply machine learning in the field of seismic data processing (Araya-Polo et al, 2019;Yang and Ma, 2019;Zhao et al, 2019). Among them, one of the most actively studied areas is prestack and poststack data noise attenuation.…”

Section: Introductionmentioning

confidence: 99%

Random Noise Attenuation of Sparker Seismic Oceanography Data with Machine Learning

Jun¹,

Jou²,

Kim³

et al. 2020

Preprint

View full text Add to dashboard Cite

Abstract. Seismic oceanography (SO) acquires water column reflections by seismic exploration compensating for the drawbacks of conventional physical oceanographic equipment. Most SO studies obtain data using air guns, which have relatively low-frequency bands. For higher-frequency bands at a low exploration cost, using a smaller seismic exploration system, such as a sparker source with a shorter receiver length, would be an alternative. However, the sparker source has a relatively low energy and consequently produces data with a low signal-to-noise (S / N) ratio. To solve the problem of the low S / N ratio of sparker SO data, we applied machine learning. The purpose of this study is to attenuate the random noise in the East Sea sparker SO data without distorting the true shape and amplitude of water column reflections. A denoising convolutional neural network (DnCNN) that successfully suppresses random noise in a natural image is adopted as the machine learning network architecture. One of the most important factors of machine learning is the generation of an appropriate training dataset. We have generated two different training datasets using synthetic and field data. Models trained with the different training datasets are applied to the test data, and the denoised results are quantitatively compared. The trained models are applied to the target seismic data, i.e., the East Sea sparker water column seismic reflection data, and the denoised seismic sections are evaluated. The results show that machine learning can successfully attenuate the random noise of sparker water column seismic reflection data.

show abstract

Deep learning-driven velocity model building workflow

Cited by 59 publications

References 19 publications

Deep Learning for Seismic Inverse Problems: Toward the Acceleration of Geophysical Analysis Workflows

Deep Learning for Seismic Inverse Problems: Toward the Acceleration of Geophysical Analysis Workflows

Random noise attenuation of sparker seismic oceanography data with machine learning

Random Noise Attenuation of Sparker Seismic Oceanography Data with Machine Learning

Contact Info

Product

Resources

About