Feature extraction based on the convolutional neural network for adaptive multiple subtraction

Li, Zhongxiao; Gao, Haotian

doi:10.1007/s11001-020-09409-7

Cited by 10 publications

(2 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Meles et al (2015) and da Costa integrate the Marchenko equation and seismic interferometry to predict prestack internal multiples. The third class is the machine learning methods concentrating on mining features from seismic data through well-trained neural networks like the generative adversarial network (Tao et al 2022), the U-Net network (Li et al 2021), the convolutional neural network (Li and Gao 2020;Liu et al 2022) and the deep neural network (Wang et al 2022). Once the networks are successfully trained, multiples can be removed efficiently.…”

Section: Introductionmentioning

confidence: 99%

Seismic forward modeling of acoustic surface-related order-separated multiples

Wang

et al. 2023

Mar Geophys Res

View full text Add to dashboard Cite

Seismic surface-related multiples have become a hot topic of great significance due to the buried geological information provided by broader illumination areas than primaries. In recent years, researchers attempt to extract the hidden hint of multiples rather than treating them as noise and eliminating them directly. The elimination methods, e.g., the surface-related multiple elimination (SRME) and the inverse scattering series free-surface multiple elimination (ISS-FSME), may be affected by the overlapping or proximity of primaries and multiples. Typical imaging methods, e.g., the reverse time migration (RTM) and the least-square reverse time migration (LSRTM), suffer severe crosstalk artifacts from multiples of inappropriate order and smooth migration velocities. To study the characteristics of primaries and surface-related multiples, whether for elimination or imaging, we propose a forward modeling method of acoustic surface-related order-separated multiples established on the areal/virtual source assumption. The free surface is replaced with an absorbing surface under the dipole source approximation and the ghost creation approach. We present two reflection operators to approximate the reflection at the free surface and apply them to the areal source to obtain ideal results. Numerical experiments on three models prove the effectiveness of the proposed forward modeling method of acoustic surface-related order-separated multiples.

show abstract

Section: Introductionmentioning

confidence: 99%

Seismic forward modeling of acoustic surface-related order-separated multiples

Wang

et al. 2023

Mar Geophys Res

View full text Add to dashboard Cite

show abstract

“…Several studies have made some attempts at multi-channel inversion of seismic signals (Li et al, 2016;Li and Gao, 2020). Zhang et al (2012) introduced a method for segmented constrained velocity inversion, that automatically selects the constrained strength and constrained range according to correlation between seismic traces, thus making the inversion results correlate better with the actual geological conditions and being more computationally efficient.…”

mentioning

confidence: 99%

Prediction of Shale Gas Pressure based on Multi‐channel Seismic Inversion in Fuling

Xing

et al. 2022

Acta Geologica Sinica (Eng)

View full text Add to dashboard Cite

The accurate prediction of formation pressure is important in oil/gas exploration and development. However, the achievement of this goal remains challenging, due to insufficient logging data and the low predictive data accuracy from seismic data. In this work, a case study was carried out in the Baima area of Wulong, in order to develop a workflow for accurately predicting shale gas formation pressure. The multi-channel stack method was first used, as well as the inversion of single-channel seismic data, to construct velocity and density models of the formation. Combined with the existing welllogging data, the velocity and density models of the whole well section were established. The shale gas formation pressure was then estimated using the Eaton method. The results show that the multi-channel seismic stacking method has a higher accuracy than the inversion of the formation velocity obtained by the single-channel seismic method. The discrepancies between our predicted formation pressure and the actual formation pressure measurement are within an acceptable range, indicating that our workflow is effective.

show abstract

Comparisons of five least-squares adaptive matched filtering methods in multiple suppression

Wang

et al. 2022

Journal of Geophysics and Engineering

View full text Add to dashboard Cite

In this paper, we comprehensively compare the application effects of five least-squares adaptive matched filtering methods (the single-channel, the multichannel, the equipoise multichannel, the pseudo multichannel and the equipoise pseudo multichannel) for multiple suppression in three representative datasets with different degrees of orthogonality. By introducing an error function, we can quantitatively analyse the influence of the five methods for multiple suppression in terms of the filter length, the normalized regularization factor, the number of matched channels, the iteration number, the amplitude ratio and noise immunity. In addition, we provide the corresponding optimal parameters or their selection principles. The comparison results show that: (i) the dependence on orthogonality is not the same for these five methods; only the equipoise multichannel and the equipoise pseudo multichannel methods can effectively reduce the dependence on orthogonality; (ii) the single-channel method is relatively balanced in all aspects; (iii) the pseudo multichannel and the equipoise pseudo multichannel methods have a stronger shaping ability but generate larger errors; (iv) the multichannel method requires a higher degree of orthogonality and (v) the optimal parameters derived from the three datasets will be better reference values for complex models or field data for the multiple suppression.

show abstract

Feature extraction based on the convolutional neural network for adaptive multiple subtraction

Cited by 10 publications

References 30 publications

Seismic forward modeling of acoustic surface-related order-separated multiples

Seismic forward modeling of acoustic surface-related order-separated multiples

Prediction of Shale Gas Pressure based on Multi‐channel Seismic Inversion in Fuling

Comparisons of five least-squares adaptive matched filtering methods in multiple suppression

Contact Info

Product

Resources

About