• Geophysics and Geophysical Exploration
• /
• v.24 no.2
• /
• pp.53-66
• /
• 2021

Velocity model building is an essential procedure in seismic data processing. Conventional techniques, such as traveltime tomography or velocity analysis take longer computational time to predict a single velocity model and the quality of the inversion results is highly dependent on human expertise. Full-waveform inversions also depend on an accurate initial model. Recently, deep neural network techniques are gaining widespread acceptance due to an increase in their integration to solving complex and nonlinear problems. This study investigated cases of seismic velocity model building using deep neural network techniques by classifying items according to the neural networks used in each study. We also included cases of generating training synthetic velocity models. Deep neural networks automatically optimize model parameters by training neural networks from large amounts of data. Thus, less human interaction is involved in the quality of the inversion results compared to that of conventional techniques and the computational cost of predicting a single velocity model after training is negligible. Additionally, unlike full-waveform inversions, the initial velocity model is not required. Several studies have demonstrated that deep neural network techniques achieve outstanding performance not only in computational cost but also in inversion results. Based on the research results, we analyzed and discussed the characteristics of deep neural network techniques for building velocity models.

• Geophysics and Geophysical Exploration
• /
• v.26 no.2
• /
• pp.84-93
• /
• 2023

The supervised learning-based deep-learning seismic inversion techniques have demonstrated successful performance in synthetic data examples targeting small-scale areas. The supervised learning-based deep-learning seismic inversion uses time-domain wavefields as input and subsurface velocity models as output. Because the time-domain wavefields contain various types of wave information, the data size is considerably large. Therefore, research applying supervised learning-based deep-learning seismic inversion trained with a significant amount of field-scale data has not yet been conducted. In this study, we predict subsurface velocity models using Laplace-domain wavefields as input instead of time-domain wavefields to apply a supervised learning-based deep-learning seismic inversion technique to field-scale data. Using Laplace-domain wavefields instead of time-domain wavefields significantly reduces the size of the input data, thereby accelerating the neural network training, although the resolution of the results is reduced. Additionally, a large grid interval can be used to efficiently predict the velocity model of the field data size, and the results obtained can be used as the initial model for subsequent inversions. The neural network is trained using only synthetic data by generating a massive synthetic velocity model and Laplace-domain wavefields of the same size as the field-scale data. In addition, we adopt a towed-streamer acquisition geometry to simulate a marine seismic survey. Testing the trained network on numerical examples using the test data and a benchmark model yielded appropriate background velocity models.

• Proceedings of the KSME Conference
• /
• 2008.11a
• /
• pp.1643-1646
• /
• 2008

In this study, an elastic network model is established in order to find dominant factors which reflect thermostability of protein structures. The connections in the elastic network model are selected with respect to the free energy between alpha-carbons, which is representatives of residues in the elastic network model. We carried out normal mode analysis and compared eigenvalues of the stiffness matrix from the elastic network model. In most cases, thermophilic proteins are observed to have higher values of lowest natural frequency than mesophiles and psychrophiles have. As a result, the thermophiles are calculated to be stiffer than other proteins in view of dynamic vibration.

• Geophysics and Geophysical Exploration
• /
• v.25 no.2
• /
• pp.71-84
• /
• 2022

When acquiring seismic data, various types of simultaneously recorded seismic noise hinder accurate interpretation. Therefore, it is essential to attenuate this noise during the processing of seismic data and research on seismic noise attenuation. For this purpose, machine learning is extensively used. This study attempts to attenuate noise in prestack seismic data using unsupervised machine learning. Three unsupervised machine learning models, N2NUNET, PATCHUNET, and DDUL, are trained and applied to synthetic and field prestack seismic data to attenuate the noise and leave clean seismic data. The results are qualitatively and quantitatively analyzed and demonstrated that all three unsupervised learning models succeeded in removing seismic noise from both synthetic and field data. Of the three, the N2NUNET model performed the worst, and the PATCHUNET and DDUL models produced almost identical results, although the DDUL model performed slightly better.

• Computational Structural Engineering
• /
• v.11 no.3
• /
• pp.241-252
• /
• 1998

본 논문은 신경망 근사 해석 모델의 원형을 스터브 거더의 거동 해석에 적용하고, 이 과정 중에 발생한 문제점을 파악하여 해결책을 제시함으로써, 앞서 개발한 원형 모델을 스터브 거더 시스템에 적합하도록 발전시키는데 목적이 있다. 스터브 거더의 해석 변수는 주어진 시간 내에 시뮬레이션이 가능하게 7개, 해석 결과값은 탄성 처짐뿐만 아니라 응력까지 고려하여 총 4개의 결과값을 동시에 고려하고, 학습 패턴 수는 총 143개를 사용하였다. 근사해석의 정확도를 향상시키고 학습의 수렴성을 보장하기 위하여 다양한 시뮬레이션을 수행하여 은닉층 뉴런 수, 학습 패턴 그리고 최대 에러의 관계를 규명하고, 이 결과를 바탕으로 신경망 근사 해석 모델 개발 단계를 수정하여 제안하였다.

• Geophysics and Geophysical Exploration
• /
• v.25 no.4
• /
• pp.227-241
• /
• 2022

Full waveform inversion (FWI) in the field of seismic data processing is an inversion technique that is used to estimate the velocity model of the subsurface for oil and gas exploration. Recently, deep learning (DL) technology has been increasingly used for seismic data processing, and its combination with FWI has attracted remarkable research efforts. For example, DL-based data processing techniques have been utilized for preprocessing input data for FWI, enabling the direct implementation of FWI through DL technology. DL-based FWI can be divided into the following methods: pure data-based, physics-based neural network, encoder-decoder, reparameterized FWI, and physics-informed neural network. In this review, we describe the theory and characteristics of the methods by systematizing them in the order of advancements. In the early days of DL-based FWI, the DL model predicted the velocity model by preparing a large training data set to adopt faithfully the basic principles of data science and apply a pure data-based prediction model. The current research trend is to supplement the shortcomings of the pure data-based approach using the loss function consisting of seismic data or physical information from the wave equation itself in deep neural networks. Based on these developments, DL-based FWI has evolved to not require a large amount of learning data, alleviating the cycle-skipping problem, which is an intrinsic limitation of FWI, and reducing computation times dramatically. The value of DL-based FWI is expected to increase continually in the processing of seismic data.

• Geophysics and Geophysical Exploration
• /
• v.23 no.3
• /
• pp.192-207
• /
• 2020

We acquire seismic data with regularly or irregularly missing traces, due to economic, environmental, and mechanical problems. Since these missing data adversely affect the results of seismic data processing and analysis, we need to reconstruct the missing data before subsequent processing. However, there are economic and temporal burdens to conducting further exploration and reconstructing missing parts. Many researchers have been studying interpolation methods to accurately reconstruct missing data. Recently, various machine learning technologies such as support vector regression, autoencoder, U-Net, ResNet, and generative adversarial network (GAN) have been applied in seismic data interpolation. In this study, by reviewing these studies, we found that not only neural network models, but also support vector regression models that have relatively simple structures can interpolate missing parts of seismic data effectively. We expect that future research can improve the interpolation performance of these machine learning models by using open-source field data, data augmentation, transfer learning, and regularization based on conventional interpolation technologies.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2008.11a
• /
• pp.672-674
• /
• 2008

• Geophysics and Geophysical Exploration
• /
• v.2 no.4
• /
• pp.167-173
• /
• 1999

Since the resolution of the 2-D hole-to-hole seismic traveltime tomography is affected by the limited ray transmission angle, various methods were used to improve the resolution. Linear traveltime interpolation(LTI) ray tracing method was chosen for forward-modeling method. Inversion results using the LTI method were compared with those using the other ray tracing methods. As an inversion algorithm, SIRT method was used. In the iterative non-linear inversion method, the cost of ray tracing is quite expensive. To reduce the cost, each raypath was stored and the inversion was performed from this information. Using the proposed method, fast convergence was achieved. Inversion results are likely to be affected by the initial velocity guess, especially when the ray transmission angle was limited. To provide a good initial guess for the inversion, generalized regression neural network(GRNN) method was used. When the transmitted raypath angle is not limited or the geological model is very complex, the inversion results are not affected by initial velocity model very much. Since the raypath angles, however, are limited in most geophysical tomographic problems, the enhancement of resolution in tomography can be achieved by providing a proper initial velocity model by another inversion algorithm such as GRNN.

• Journal of Korean Tunnelling and Underground Space Association
• /
• v.14 no.1
• /
• pp.37-53
• /
• 2012

There is considerable uncertainty in ground properties used in tunnel designs. In this study, a back analysis was performed to find optimal ground properties based on the artificial neural network facility of MATLAB program of using tunnel monitoring data. Total 81 data were constructed by changing elastic modulus and coefficient of lateral pressure which have great influence on tunnel convergence. A sensitivity analysis was conducted to establish an optimal training model by varying the number of hidden layers, the number of nodes, learning rate, and momentum. Meanwhile, the optimal training model was selected by comparing MSE (Mean Squared Error) and coefficient of determination ($R^2$) and was used to find the correct elastic moduli of layers and the coefficient of lateral pressure. In future, it is expected that the suggested method of this study can be applied to determine the optimum tunnel support pattern under given ground conditions.