Regardless of successful applications the convolutional neural networks (CNNs) in different fields, its application to seismic waveform classification and first-break (FB) picking has not been explored yet. This letter investigates CNNs for classifying time-space waveforms from shot gathers FBs both direct wave refracted wave. We use representative subimage samples with two types labeled supervise training. The goal is obtain optimal weights biases CNNs, which are solved by minimizing error...

The impedance inversion technique plays a crucial role in seismic reservoir properties prediction. However, most existing methods often suffer from spatial discontinuities and instability because each vertical profile is processed independently the inversion. We tested transform-domain sparsity promotion simultaneous multitrace method to address this issue. approach was implemented through minimizing data misfit term constraint that incorporates (2D or 3D) structural information into...

In seismic exploration, the wavelet-filtering effect and Q-filtering (amplitude attenuation velocity dispersion) blur reflection image of subsurface layers. Therefore, both wavelet- effects should be reduced to retrieve a high-quality image, which is significant for fine reservoir interpretation. We derive nonlinear time-variant convolution model sparsely represent nonstationary seismograms in time domain involving these two present deconvolution (TVD) method based on sparse Bayesian...

Conventional reflectivity inversion methods are based on a stationary convolution model and theoretically require seismic traces as input (i.e., those free of attenuation dispersion effects). Reflectivity for nonstationary data, which is typical field surveys, requires us to first compensate the earth’s [Formula: see text]-filtering effects by inverse text] filtering. However, compensation filtering inherently unstable, offers no perfect solution. Thus, we presented sparse method data. We...

Prestack acoustic full-waveform inversion (FWI) can provide long-wavelength components of the P-wave velocity by using low frequencies and long-offset direct/diving/refracted waves, which could be simulated via a large space grid, it is weakly sensitive to density. Poststack impedance usually quickly yield high-resolution impedance, Therefore, we have combined these two methods develop an FWI-driven inversion. Our method first uses FWI obtain with guaranteed overlap between high poststack...

ABSTRACT A spectral sparse Bayesian learning reflectivity inversion method, combining with learning, is presented in this paper. The method retrieves a series by sequentially adding, deleting or re‐estimating hyper‐parameters, without pre‐setting the number of non‐zero spikes. spikes largest amplitude are usually first to be resolved. tested on data sets, including synthetic data, physical modelling and field sets. results show that can identify thin beds below tuning thickness highlight...

Abstract We developed a system to explore the effects of pressure and fluid viscosity on dispersion attenuation fully saturated tight sandstones, especially at seismic frequencies. Calibration new revealed that can operate reliably frequencies [2–200, 10 6 ] Hz. Tight sandstone with “crack–pore” microstructure was tested under nitrogen gas (dry), brine, glycerin saturation. A frequency‐dependent effect not found for dry case. However, apparent undrained/unrelaxed transition clearly observed...

Physical and/or economic constraints cause acquired seismic data to be incomplete; however, complete are required for many subsequent processing procedures. Data reconstruction is a crucial and long-standing topic in the exploration seismology field. We extended our previous works on deep learning (DL)-based irregularly regularly missing 2-D 3-D data. A key motivation that convolutional neural network (CNN) can take full advantage of nature data, additional dimension allows more information...

Inspired by image-to-image translation, we applied deep learning (DL) to regularly missing data reconstruction, aimed at translating incomplete into their corresponding complete data. With this purpose in mind, first construct a network architecture based on an end-to-end U-Net convolutional network, which is generic DL solution for various tasks. We then meticulously prepare the training with both synthetic and field seismic This article implemented Python Keras (a high-level library)....

Low-frequency information is important in reducing the nonuniqueness of absolute impedance inversion and for quantitative seismic interpretation. In traditional model-driven methods, low-frequency background from an initial model almost unchanged during process. Moreover, results are limited by quality modeled data extracted wavelet. To alleviate these issues, we have investigated a double-scale supervised method based on gated recurrent encoder-decoder network (GREDN). We first train...

Nonstationary seismic data can be expressed using a linear matrix-vector multiplication system derived from wave theory when anelastic effects of the earth quantified by intrinsic quality factor (or [Formula: see text]) and text] is frequency independent in bandwidth. On basis modeling singular value decomposition, we have assessed stability weights associated with left vectors, data, values, compensation/resolution limitation inversion-based deabsorption right vectors. In addition, stable...

Traditional 1-D instantaneous phase (IP) is a routine attribute for detecting structural discontinuities of seismic images. The has the ability to detect subtle changes, but it meanwhile sensitive noise. Furthermore, traditional IP calculated trace by and thus cannot effectively utilize geological constraints. sensitivity noise unavailability constraints limit practical applications attributes. To address these two issues, this letter proposes 3-D geosteering derived from IP. At first, we...

Seismic data are increasingly required to be high quality for the continuous improvement of degree exploration. From viewpoint inversion, utilization more information is an effective way improve signal-to-noise ratio seismic data. In this letter, we adopt simultaneous sparsity constraints first-order differences signals along time direction and two spatial directions, described by minimizing Cauchy function, as a combined constraint (or regularization) term imposed on time-domain misfit...

Noise attenuation has been a long-standing but still active topic in seismic data processing. The deep convolutional neural networks (CNNs) have recently adopted to remove the learned random noise from noisy data, it is difficult improve generalization ability of denoisers due limited diversity training sets. In this letter, we investigate an end-to-end denoising CNNs (DCNNs) with novel generation method involving multidimensional geological structure features for denoising. To learn...

Attenuating random noise while preserving edges and structures of seismic signals is great significance for interpretation inversion. An edge-preserving signal-preserving reduction method presented in this paper. The regards problem as an inverse based on a Bayesian framework. A priori information that the difference data obey Cauchy distribution, kind non-Gaussian used constraint to control reduction. performance mainly dependent trade-off parameter scale parameter, which make it easier...

Time-frequency (TF) analysis is a useful tool for seismic data processing and interpretation. We introduce sparse Bayesian learning (SBL) to TF propose new SBL-based high-resolution method. The method decomposes the trace into series of Ricker wavelets using representations subsequently implements Wigner-Ville distribution (WVD) on decomposed produce spectra. By iteratively solving maximum posterior type-II likelihood, decomposition can sequentially obtain an optimal number with different...

Previous studies have demonstrated that P-wave velocity dispersion at seismic frequencies is often related to hydrocarbons, which results in frequency-dependent reflection coefficients. This effect neglected the conventional amplitude-versus-angle (AVA) inversion, or reduced most AVA inversion involving due linearization of either forward modeling objective function. As a consequence, there are times when nonnegligible error exists inverted dispersion-associated result, probably some cases....

P- and S-wave velocity attenuation coefficients (accurate to ±0.3% ±0.2 dB/cm, respectively) were measured in synthetic porous rocks with aligned, penny-shaped fractures using the laboratory ultrasonic pulse-echo method. Shear-wave splitting was observed by rotating transducer noting maximum minimum velocities relative fracture direction. A block of rock density 0.0201 ± 0.0068 size 3.6 0.38 mm (measured from image analysis X-ray CT scans) sub-sampled into three 20–30 long, 50 diameter core...

Enlightened by the classical total variation (TV) model, we present a novel random noise reduction method for seismic data based on Bayesian inversion, called inversion filtering. The regards 2D 'clean' as model parameters, and is equivalent to inverting these parameters from observed data. implemented maximizing posterior distribution, which replaced product of priori distribution likelihood function. performance this mainly depends choice information. Based statistical knowledge or...

We have developed a method of nonstationary sparse reflectivity inversion (NSRI) that directly retrieves the series from seismic data without intrinsic instability associated with inverse [Formula: see text] filtering methods. investigated NSRI performance in presence input error (e.g., phase shift and peak frequency wavelet), which determined results are reasonable case moderate error. was then applied to collected laboratory physical model made highly attenuating media, for true...