Deep learning-driven velocity model building workflow

2019 ◽  
Vol 38 (11) ◽  
pp. 872a1-872a9 ◽  
Author(s):  
Mauricio Araya-Polo ◽  
Stuart Farris ◽  
Manuel Florez
Keyword(s):  
Deep Learning ◽  
Seismic Data ◽  
Model Building ◽  
Ground Truth ◽  
Velocity Model ◽  
Training Data ◽  
Quality Data ◽  
Generative Adversarial Network ◽  
Data Set ◽  
Velocity Models

Exploration seismic data are heavily manipulated before human interpreters are able to extract meaningful information regarding subsurface structures. This manipulation adds modeling and human biases and is limited by methodological shortcomings. Alternatively, using seismic data directly is becoming possible thanks to deep learning (DL) techniques. A DL-based workflow is introduced that uses analog velocity models and realistic raw seismic waveforms as input and produces subsurface velocity models as output. When insufficient data are used for training, DL algorithms tend to overfit or fail. Gathering large amounts of labeled and standardized seismic data sets is not straightforward. This shortage of quality data is addressed by building a generative adversarial network (GAN) to augment the original training data set, which is then used by DL-driven seismic tomography as input. The DL tomographic operator predicts velocity models with high statistical and structural accuracy after being trained with GAN-generated velocity models. Beyond the field of exploration geophysics, the use of machine learning in earth science is challenged by the lack of labeled data or properly interpreted ground truth, since we seldom know what truly exists beneath the earth's surface. The unsupervised approach (using GANs to generate labeled data)illustrates a way to mitigate this problem and opens geology, geophysics, and planetary sciences to more DL applications.

SaltSeg: Automatic 3D salt segmentation using a deep convolutional neural network

2019 ◽  
Vol 7 (3) ◽  
pp. SE113-SE122 ◽  
Author(s):  
Yunzhi Shi ◽  
Xinming Wu ◽  
Sergey Fomel
Keyword(s):  
Large Scale ◽  
Model Building ◽  
Ground Truth ◽  
Velocity Model ◽  
Training Data ◽  
Data Sets ◽  
Validation Data ◽  
Data Set ◽  
Seismic Image ◽  
Data Generator

Salt boundary interpretation is important for the understanding of salt tectonics and velocity model building for seismic migration. Conventional methods consist of computing salt attributes and extracting salt boundaries. We have formulated the problem as 3D image segmentation and evaluated an efficient approach based on deep convolutional neural networks (CNNs) with an encoder-decoder architecture. To train the model, we design a data generator that extracts randomly positioned subvolumes from large-scale 3D training data set followed by data augmentation, then feed a large number of subvolumes into the network while using salt/nonsalt binary labels generated by thresholding the velocity model as ground truth labels. We test the model on validation data sets and compare the blind test predictions with the ground truth. Our results indicate that our method is capable of automatically capturing subtle salt features from the 3D seismic image with less or no need for manual input. We further test the model on a field example to indicate the generalization of this deep CNN method across different data sets.

Seismic Velocity Model Building Using Neural Networks: Training Data Design and Learning Generalization

Geophysics ◽  
2021 ◽  
pp. 1-73
Author(s):  
Hani Alzahrani ◽  
Jeffrey Shragge
Keyword(s):  
Seismic Data ◽  
Model Building ◽  
Seismic Velocity ◽  
Velocity Model ◽  
Training Data ◽  
Training Process ◽  
Training Models ◽  
Velocity Models ◽  
Velocity Model Building ◽  
Layer Velocity

Data-driven artificial neural networks (ANNs) offer a number of advantages over conventional deterministic methods in a wide range of geophysical problems. For seismic velocity model building, judiciously trained ANNs offer the possibility of estimating high-resolution subsurface velocity models. However, a significant challenge of ANNs is training generalization, which is the ability of an ANN to apply the learning from the training process to test data not previously encountered. In the context of velocity model building, this means learning the relationship between velocity models and the corresponding seismic data from a set of training data, and then using acquired seismic data to accurately estimate unknown velocity models. We ask the following question: what type of velocity model structures need be included in the training process so that the trained ANN can invert seismic data from a different (hypothetical) geological setting? To address this question, we create four sets of training models: geologically inspired and purely geometrical, both with and without background velocity gradients. We find that using geologically inspired training data produce models with well-delineated layer interfaces and fewer intra-layer velocity variations. The absence of a certain geological structure in training models, though, hinders the ANN's ability to recover it in the testing data. We use purely geometric training models consisting of square blocks of varying size to demonstrate the ability of ANNs to recover reasonable approximations of flat, dipping, and curved interfaces. However, the predicted models suffer from intra-layer velocity variations and non-physical artifacts. Overall, the results successfully demonstrate the use of ANNs in recovering accurate velocity model estimates, and highlight the possibility of using such an approach for the generalized seismic velocity inversion problem.

scholarly journals Deep-learning inversion: A next-generation seismic velocity model building method

Geophysics ◽  
2019 ◽  
Vol 84 (4) ◽  
pp. R583-R599 ◽  
Author(s):  
Fangshu Yang ◽  
Jianwei Ma
Keyword(s):  
Deep Learning ◽  
Seismic Data ◽  
Model Building ◽  
Seismic Velocity ◽  
Velocity Model ◽  
Computational Time ◽  
Learning Methods ◽  
Velocity Models ◽  
Velocity Model Building ◽  
Training Stage

Seismic velocity is one of the most important parameters used in seismic exploration. Accurate velocity models are the key prerequisites for reverse time migration and other high-resolution seismic imaging techniques. Such velocity information has traditionally been derived by tomography or full-waveform inversion (FWI), which are time consuming and computationally expensive, and they rely heavily on human interaction and quality control. We have investigated a novel method based on the supervised deep fully convolutional neural network for velocity-model building directly from raw seismograms. Unlike the conventional inversion method based on physical models, supervised deep-learning methods are based on big-data training rather than prior-knowledge assumptions. During the training stage, the network establishes a nonlinear projection from the multishot seismic data to the corresponding velocity models. During the prediction stage, the trained network can be used to estimate the velocity models from the new input seismic data. One key characteristic of the deep-learning method is that it can automatically extract multilayer useful features without the need for human-curated activities and an initial velocity setup. The data-driven method usually requires more time during the training stage, and actual predictions take less time, with only seconds needed. Therefore, the computational time of geophysical inversions, including real-time inversions, can be dramatically reduced once a good generalized network is built. By using numerical experiments on synthetic models, the promising performance of our proposed method is shown in comparison with conventional FWI even when the input data are in more realistic scenarios. We have also evaluated deep-learning methods, the training data set, the lack of low frequencies, and the advantages and disadvantages of our method.

Diffraction imaging and depth-velocity inversion with 3D P-Cable seismic data

2021 ◽  
Author(s):  
Alexander Bauer ◽  
Benjamin Schwarz ◽  
Dirk Gajewski
Keyword(s):  
Seismic Data ◽  
Model Building ◽  
Computational Cost ◽  
Velocity Model ◽  
Data Cube ◽  
Velocity Inversion ◽  
Velocity Models ◽  
North East ◽  
3D Velocity Model ◽  
Zero Offset

<p>Most established methods for the estimation of subsurface velocity models rely on the measurements of reflected or diving waves and therefore require data with sufficiently large source-receiver offsets. For seismic data that lacks these offsets, such as vintage data, low-fold academic data or near zero-offset P-Cable data, these methods fail. Building on recent studies, we apply a workflow that exploits the diffracted wavefield for depth-velocity-model building. This workflow consists of three principal steps: (1) revealing the diffracted wavefield by modeling and adaptively subtracting reflections from the raw data, (2) characterizing the diffractions with physically meaningful wavefront attributes, (3) estimating depth-velocity models with wavefront tomography. We propose a hybrid 2D/3D approach, in which we apply the well-established and automated 2D workflow to numerous inlines of a high-resolution 3D P-Cable dataset acquired near Ritter Island, a small volcanic island located north-east of New Guinea known for a catastrophic flank collapse in 1888. We use the obtained set of parallel 2D velocity models to interpolate a 3D velocity model for the whole data cube, thus overcoming possible issues such as varying data quality in inline and crossline direction and the high computational cost of 3D data analysis. Even though the 2D workflow may suffer from out-of-plane effects, we obtain a smooth 3D velocity model that is consistent with the data.</p>

Velocity model building in a crosswell acquisition geometry with image-trained artificial neural networks

Geophysics ◽  
2020 ◽  
Vol 85 (2) ◽  
pp. U31-U46
Author(s):  
Wenlong Wang ◽  
Jianwei Ma
Keyword(s):  
Wave Equations ◽  
Model Building ◽  
Velocity Model ◽  
P Wave ◽  
Common Source ◽  
Data Set ◽  
Velocity Models ◽  
Velocity Model Building ◽  
Artificial Neural ◽  
The Cost

We have developed an artificial neural network to estimate P-wave velocity models directly from prestack common-source gathers. Our network is composed of a fully connected layer set and a modified fully convolutional layer set. The parameters in the network are tuned through supervised learning to map multishot common-source gathers to velocity models. To boost the generalization ability, the network is trained on a massive data set in which the velocity models are modified from natural images that are collected from an online repository. Multishot seismic traces are simulated from those models with acoustic wave equations in a crosswell acquisition geometry. Shot gathers from different source positions are transformed as channels in the network to increase data redundancy. The training process is expensive, but it only occurs once up front. The cost for predicting velocity models is negligible once the training is complete. Different variations of the network are trained and analyzed. The trained networks indicate encouraging results for predicting velocity models from prestack seismic data that are acquired with the same geometry as in the training set.

scholarly journals Inspiration for Seismic Diffraction Modelling, Separation, and Velocity in Depth Imaging

2020 ◽  
Vol 10 (12) ◽  
pp. 4391
Author(s):  
Yasir Bashir ◽  
Nordiana Mohd Muztaza ◽  
Seyed Yaser Moussavi Alashloo ◽  
Syed Haroon Ali ◽  
Deva Prasad Ghosh
Keyword(s):  
Travel Time ◽  
Seismic Data ◽  
Oil And Gas ◽  
Model Building ◽  
Research Work ◽  
Velocity Model ◽  
Small Scale ◽  
Velocity Models ◽  
Analysis Methods ◽  
Impedance Contrast

Fractured imaging is an important target for oil and gas exploration, as these images are heterogeneous and have contain low-impedance contrast, which indicate the complexity in a geological structure. These small-scale discontinuities, such as fractures and faults, present themselves in seismic data in the form of diffracted waves. Generally, seismic data contain both reflected and diffracted events because of the physical phenomena in the subsurface and due to the recording system. Seismic diffractions are produced once the acoustic impedance contrast appears, including faults, fractures, channels, rough edges of structures, and karst sections. In this study, a double square root (DSR) equation is used for modeling of the diffraction hyperbola with different velocities and depths of point diffraction to elaborate the diffraction hyperbolic pattern. Further, we study the diffraction separation methods and the effects of the velocity analysis methods (semblance vs. hybrid travel time) for velocity model building for imaging. As a proof of concept, we apply our research work on a steep dipping fault model, which demonstrates the possibility of separating seismic diffractions using dip frequency filtering (DFF) in the frequency–wavenumber (F-K) domain. The imaging is performed using two different velocity models, namely the semblance and hybrid travel time (HTT) analysis methods. The HTT method provides the optimum results for imaging of complex structures and imaging below shadow zones.

Frequency and spatial sampling strategies for crosshole seismic waveform spectral inversion experiments

Geophysics ◽  
2009 ◽  
Vol 74 (6) ◽  
pp. WCC79-WCC89 ◽  
Author(s):  
Hansruedi Maurer ◽  
Stewart Greenhalgh ◽  
Sabine Latzel
Keyword(s):  
Information Content ◽  
Seismic Data ◽  
Velocity Model ◽  
Similar Amount ◽  
Data Sets ◽  
Data Set ◽  
Waveform Data ◽  
Velocity Models ◽  
Seismic Waveform ◽  
Iterative Inversion

Analyses of synthetic frequency-domain acoustic waveform data provide new insights into the design and imaging capability of crosshole surveys. The full complex Fourier spectral data offer significantly more information than other data representations such as the amplitude, phase, or Hartley spectrum. Extensive eigenvalue analyses are used for further inspection of the information content offered by the seismic data. The goodness of different experimental configurations is investigated by varying the choice of (1) the frequencies, (2) the source and receiver spacings along the boreholes, and (3) the borehole separation. With only a few carefully chosen frequencies, a similar amount of information can be extracted from the seismic data as can be extracted with a much larger suite of equally spaced frequencies. Optimized data sets should include at least one very low frequencycomponent. The remaining frequencies should be chosen fromthe upper end of the spectrum available. This strategy proved to be applicable to a simple homogeneous and a very complex velocity model. Further tests are required, but it appears on the available evidence to be model independent. Source and receiver spacings also have an effect on the goodness of an experimental setup, but there are only minor benefits to denser sampling when the increment is much smaller than the shortest wavelength included in a data set. If the borehole separation becomes unfavorably large, the information content of the data is degraded, even when many frequencies and small source and receiver spacings are considered. The findings are based on eigenvalue analyses using the true velocity models. Because under realistic conditions the true model is not known, it is shown that the optimized data sets are sufficiently robust to allow the iterative inversion schemes to converge to the global minimum. This is demonstrated by means of tomographic inversions of several optimized data sets.

Ground-roll attenuation using generative adversarial networks

Geophysics ◽  
2020 ◽  
Vol 85 (4) ◽  
pp. WA255-WA267 ◽  
Author(s):  
Yijun Yuan ◽  
Xu Si ◽  
Yue Zheng
Keyword(s):  
Seismic Data ◽  
Signal To Noise Ratio ◽  
Training Data ◽  
Generative Adversarial Networks ◽  
Coherent Noise ◽  
Traditional Methods ◽  
Generative Adversarial Network ◽  
Data Set ◽  
Persistent Problem ◽  
Ground Roll

Ground roll is a persistent problem in land seismic data. This type of coherent noise often contaminates seismic signals and severely reduces the signal-to-noise ratio of seismic data. A variety of methods for addressing ground-roll attenuation have been developed. However, existing methods are limited, especially when using real land seismic data. For example, when ground roll and reflections overlap in the time or frequency domains, traditional methods cannot completely separate them and they often distort the signals during the suppression process. We have developed a generative adversarial network (GAN) to attenuate ground roll in seismic data. Unlike traditional methods for ground-roll attenuation dependent on various filters, the GAN method is based on a large training data set that includes pairs of data with and without ground roll. After training the neural network with the training data, the network can identify and filter out any noise in the data. To fulfill this purpose, the proposed method uses a generator and a discriminator. Through network training, the generator learns to create the data that can fool the discriminator, and the discriminator can then distinguish between the data produced by the generator and the training data. As a result of the competition between generators and discriminators, generators produce better images whereas discriminators accurately recognize targets. Tests on synthetic and real land seismic data show that the proposed method effectively reveals reflections masked by the ground roll and obtains better results in the attenuation of ground roll and in the preservation of signals compared to the three other methods.

Localized anisotropic tomography with well information in VTI media

Geophysics ◽  
2010 ◽  
Vol 75 (5) ◽  
pp. D37-D45 ◽  
Author(s):  
Andrey Bakulin ◽  
Marta Woodward ◽  
Dave Nichols ◽  
Konstantin Osypov ◽  
Olga Zdraveva
Keyword(s):  
Vertical Velocity ◽  
Seismic Data ◽  
Model Building ◽  
Synthetic Data ◽  
Resolution Limit ◽  
Data Set ◽  
Anisotropic Models ◽  
Velocity Models ◽  
First Case ◽  
Well Data

We develop a concept of localized seismic grid tomography constrained by well information and apply it to building vertically transversely isotropic (VTI) velocity models in depth. The goal is to use a highly automated migration velocity analysis to build anisotropic models that combine optimal image focusing with accurate depth positioning in one step. We localize tomography to a limited volume around the well and jointly invert the surface seismic and well data. Well information is propagated into the local volume by using the method of preconditioning, whereby model updates are shaped to follow geologic layers with spatial smoothing constraints. We analyze our concept with a synthetic data example of anisotropic tomography applied to a 1D VTI model. We demonstrate four cases of introducing additionalinformation. In the first case, vertical velocity is assumed to be known, and the tomography inverts only for Thomsen’s [Formula: see text] and [Formula: see text] profiles using surface seismic data alone. In the second case, tomography simultaneously inverts for all three VTI parameters, including vertical velocity, using a joint data set that consists of surface seismic data and vertical check-shot traveltimes. In the third and fourth cases, sparse depth markers and walkaway vertical seismic profiling (VSP) are used, respectively, to supplement the seismic data. For all four examples, tomography reliably recovers the anisotropic velocity field up to a vertical resolution comparable to that of the well data. Even though walkaway VSP has the additional dimension of angle or offset, it offers no further increase in this resolution limit. Anisotropic tomography with well constraints has multiple advantages over other approaches and deserves a place in the portfolio of model-building tools.

scholarly journals A deep learning segmentation strategy that minimizes the amount of manually annotated images

F1000Research ◽  
2021 ◽  
Vol 10 ◽  
pp. 256
Author(s):  
Thierry Pécot ◽  
Alexander Alekseyenko ◽  
Kristin Wallace
Keyword(s):  
Deep Learning ◽  
Data Augmentation ◽  
Training Data ◽  
Artificial Data ◽  
Deep Convolutional Neural Networks ◽  
Generative Adversarial Network ◽  
Data Set ◽  
Segmentation Strategy ◽  
Adversarial Network ◽  
The Impact

Deep learning has revolutionized the automatic processing of images. While deep convolutional neural networks have demonstrated astonishing segmentation results for many biological objects acquired with microscopy, this technology's good performance relies on large training datasets. In this paper, we present a strategy to minimize the amount of time spent in manually annotating images for segmentation. It involves using an efficient and open source annotation tool, the artificial increase of the training data set with data augmentation, the creation of an artificial data set with a conditional generative adversarial network and the combination of semantic and instance segmentations. We evaluate the impact of each of these approaches for the segmentation of nuclei in 2D widefield images of human precancerous polyp biopsies in order to define an optimal strategy.

Sign in / Sign up

Export Citation Format

Share Document