Kirchhoff Migration Research Articles

Self-adaptive denoising net: Self-supervised learning for seismic migration artifacts and random noise attenuation

Seismic noise attenuation is essential for seismic interpretation and reservoir characterization. Recently, many researchers have applied convolutional neural network (CNN) to attenuate seismic noise. We propose a novel self-supervised CNN based approach to attenuate seismic random noise and migration artifacts simultaneously, termed as self-adaptive denoising net (SaDN). In this approach, we modify the loss-function of denoising CNN (DnCNN) based on the assumption that the synthetic noise with mixed Gaussian-Poisson distribution can simulate random noise and migration artifact. Moreover, we develop a new framework for the training and testing procedure. By adding the synthetic noise to seismic data, we can use the proposed CNN model to learn the characteristic of signal-dependent noise. The additive signal-dependent noise is self-adapted to the original seismic data according to the statistical property. Then, the well-trained SaDN model is applied to the original 3D seismic survey to recognize and extract the original random noise and migration artifact. The SaDN is training in the original seismic data with synthetic noise and testing in the same targeting seismic data without synthetic noise. Besides, our method does not require other denoising method for preprocessing in the training data. Thus, our proposed SaDN can be trained in a self-supervised way and adapted to different field data. Different experiments have been tested and compare with the method of CWT to illustrate the robustness and superiority of our proposed SaDN. The experiment results indicate that our proposed SaDN is able to remove the seismic random noise and migration artifacts without harming the effective signal.

Triggered Seismicity in a Flooded Former Coal Mining Basin (Gardanne Area, France)

We studied the correlation between seismicity and the water table level in an abandoned coal mine (Bouches-du-Rhône, France), closed in 2003, where groundwater has been pumped out since 2010 to prevent underground flooding. Microseismicity was first felt by the population in 2010 and a strongly felt seismic swarm occurred in November 2012. The origin of the seismicity was therefore questioned, in relation to both the potential instability of old, shallow galleries that might generate damage at the surface and a local seismic hazard assessment. A temporary dense seismic network in the area allowed us to analyse the spatial distribution of the seismicity in detail. Most of the seismicity was clearly located under the mine workings, highlighting that an existing fault system crossing the mining operation was being hydraulically activated, in accordance with the known tectonic extension regime. Our analysis clearly shows a spatiotemporal relationship between seismic migration and the level of the mine aquifer between 2013 and 2017. Thus, seismicity will persist with oscillations of the mining aquifer, depending on the pumping capacities and effective rainfall. Continuous hydraulic and seismic monitoring is necessary to better understand these phenomena and assess the associated risks.

Theoretical and quantitative evaluation of hybrid PML-ABCs for seismic wave simulation

A good artificial boundary treatment in a seismic wave grid-based numerical simulation can reduce the size of the computational region and increase the computational efficiency, which is becoming increasingly important for seismic migration and waveform inversion tasks requiring hundreds or thousands of simulations. Two artificial boundary techniques are commonly used: perfectly matched layers (PMLs), which exhibit the excellent absorption performance but impose a greater computational burden by using finite layers to gradually reduce wave amplitudes; and absorbing boundary conditions (ABCs), which have the high computational efficiency but are less effective in absorption because they employ the one-way wave equation at the exterior boundary. Naturally, PMLs have been combined with ABCs to reduce the number of PMLs, thus improving the computational efficiency; many studies have proposed such hybrid PMLs. Depending on the equations from which the ABCs are derived, there are two hybrid PML variants: the PML+unstretched ABC (UABC), in which the ABC is derived from a physical equation; or the PML+stretched ABC (SABC), in which the ABC is derived from the PML equation. Even though all the previous studies concluded that hybrid PMLs can improve the absorption performance, none of them quantified how many PMLs can be removed by combining the PML with the ABC compared with the pure PML. In this paper, we systematically study the absorption performance of the two hybrid PML variants. We develop a method to distinguish the artificial reflections from the PML-interior interface and those caused by the PML exterior boundary to accurately approximate the additional absorption achieved by using the UABC and the SABC. The reflection coefficients based on a theoretical derivation and numerical tests both show that the UABC amplifies most reflections and is not recommended in any situation; conversely, the SABC can always diminish reflections, but the additional absorption achieved by the SABC is relatively poor and cannot effectively reduce the number of PMLs. In contrast, we find that simply increasing the damping parameter improves absorption better than the PML+SABC. Our results show that the improvement in absorption achieved by combining the PML with either the SABC or the UABC is not better than that obtained by simply adjusting the damping profile of the PML; thus, combining the PML with the ABC is not recommended in practice.

The Inverse Fresnel Beam XSP-CDP Stack Imaging in Crosswell Seismic

In order to overcome the shortcomings of serious arc drawing and low computational efficiency in the crosswell seismic migration method and the problems of the inaccurate velocity model and sparse distribution of reflection points in the traditional stack imaging method, the article proposes an inverse Fresnel beam XSP-CDP stack imaging method based on first-arrival wave velocity tomography combined with the characteristics of crosswell seismic wave field. Firstly, an accurate crosswell velocity model is established by the first-arrival wave tomography inversion method based on the characteristics of high energy and easy pick-up of the first-arrival wave in crosswell seismic. Secondly, the velocity model is optimized, and the energy contribution weights of effective rays to the receiver point are calculated through the crosswell seismic Fresnel beam wave field forward numerical simulation method. Then, the reflected wave field is dynamically migrated to the reflection points within the first Fresnel zone according to the weight function, and the intensive common reflection point (CRP) gather after normal moveout (NMO) correction is generated. Finally, an appropriate bin is selected for stacking. In this article, the inverse Fresnel beam method is used to decompose the single-channel seismic wave field into the effective reflection points in the Fresnel zone, which makes the fold of the reflection point more uniform and improves the imaging accuracy. The model test and actual data processing results proved the validity and robustness of this method.

Reverse time migration of phase-encoded all-order multiples

Multiples follow long paths and carry more information on the subsurface than primary reflections, making them particularly useful for imaging. However, seismic migration using multiples can generate crosstalk artifacts in the resulting images because multiples of different orders interfere with each other and crosstalk artifacts greatly degrade the quality of an image. We form a supergather by applying phase-encoding functions to image multiples and stacking several encoded controlled-order multiples. The multiples are separated into different orders using multiple decomposition strategies. The method is referred to as the phase-encoded migration of all-order multiples (PEM). The new migration can be performed by applying only two finite-difference solutions to the wave equation. The solutions include backward extrapolating the blended virtual receiver data and forward propagating the summed virtual source data. Our approach can significantly attenuate crosstalk artifacts and also significantly reduce computational costs. Numerical examples demonstrate that PEM can remove relatively strong crosstalk artifacts generated by multiples and is a promising approach for imaging subsurface targets.

A three-dimensional Laguerre one-way wave equation solver

A finite difference algorithm based on the integral Laguerre transform in time for solving a three-dimensional one-way wave equation is proposed. This allows achieving high accuracy of calculation results. In contrast to the Fourier method, the approach does not need to solve systems of linear algebraic equations with indefinite matrices. To filter the unstable components of a wave field, Richardson extrapolation or spline approximation can be used. However, these methods impose additional limitations on the integration step in depth. This problem can be solved if the filtering is performed not in the direction of extrapolation of the wave field, but in a horizontal plane. This approach called for fast methods of converting the Laguerre series coefficients into the Fourier series coefficients and vice versa. The high stability of the new algorithm allows calculations with a large depth step without loss of accuracy and, in combination with Marchuk-Strang splitting, this can significantly reduce the calculation time. Computational experiments are performed. The results have shown that this algorithm is highly accurate and efficient in solving the problems of seismic migration.

Fusion Before Imaging Method for Heterogeneous Borehole Radar Subsurface Surveys

In this article, an efficient radar fusion-before-imaging (RFBI) method for heterogeneous borehole radar systems is proposed. Different from conventional fusion-after-imaging processes, the sample sets collected from heterogeneous borehole radar systems (monostatic, bistatic, or multiple-input multiple-output) are first merged into one data set before the imaging process in RFBI, and a single imaging operation is demanded to obtain the target space image with high precision. Specifically, the diversity in heterogeneous borehole radar sample sets is taken into consideration, and the radar sample sets are inserted and fused into a high-dimensional sample set before imaging. The target space spectrum is generated according to the echo space-frequency constraint relationship, and the target space is extracted from one imaging process. The influence of clutters in RFBI results is reduced, and the imaging accuracy is satisfactory. Meanwhile, due to the fusion process ahead, the computational time of RFBI hardly increases with the number of radar sample sets. The synthetic and field experiment results show that RFBI demonstrates comparable accuracy as Kirchhoff migration and higher efficiency in processing large amounts of data sets at the cost of large memory requirement, which is suitable for the joint imaging of heterogeneous radar systems.

Feasibility of Bone Fracture Detection Using Microwave Imaging

This paper studies the feasibility of Microwave Imaging (MWI) for detection of fractures in superficial bones like the tibia, using a simple and practical setup. First-responders could use it for fast preliminary diagnosis in emergency locations, where X-Rays are not available. It may prove valuable also for cases where X-ray are not recommended, e.g., length pregnant women or children. The method is inspired on the synthetic aperture radar technique. A single Vivaldi antenna is used to linearly scan the bone in the 8.3-11.1 GHz frequency range and collect the scattered fields. The system is operated in air, without the need for impractical impedance-matching immersion liquids. The image is reconstructed using a Kirchhoff migration algorithm. A Singular Value Decomposition (SVD) strategy is used to remove skin and background artifacts. To test this technique, a set of full-wave simulations and experiments were conducted on a multilayer phantom and on an ex-vivo animal bone. Results show that the system can detect and locate bone transverse fractures as small as 1 mm width and 13 mm deep, even when the bone is wrapped by 2 mm thick skin.

Nondestructive Inspection of Tree Trunks Using a Dual-Polarized Ground-Penetrating Radar System

In recent years, trees in European countries have been increasingly endangered by emerging infectious diseases (EIDs). In the United Kingdom, this has been observed to affect whole woodlands and forests, threatening the existence of some types of trees. Although quarantine measures have been taken to limit the spreading of such diseases, this has not yet been effectively controlled leading to millions of trees affected by EIDs. Ground-penetrating radar (GPR) has proven effective in identifying critical features on diseased trees for detection of EIDs spread. However, the irregular shape of tree trunks and their complex internal structure represent real challenges for conventional GPR measurements and data processing methodologies. In this research, a dual-polarised GPR system is used to detect internal decay in tree trunks using novel signal processing methodologies. A polarimetric correlation filter based on Bragg Scattering on a 3D Pauli feature vector and an arc-shaped Kirchhoff migration are discussed in detail. The proposed polarimetric correlation filter is utilised to enhance the signal-to-noise ratio (SNR) of B-scans due to bark and tree trunk high-loss properties of tree trunks. Meanwhile, an arc-shaped Kirchhoff migration algorithm is performed to counteract the influence of the bark irregularity. The proposed data processing framework is successfully validated with measurements on a real tree trunk, where cross-sections were subsequently cut for comparison purposes. Outcomes from the proposed methodology demonstrate a high consistency with the features observed on the tree trunk cross-sections, indicating the reliability of the proposed detection scheme for assessing tree-decay associated with EIDs.

Deep-Learning Full-Waveform Inversion Using Seismic Migration Images

Data-driven deep-learning full-waveform inversion (DD-DLFWI) can efficiently reconstruct a velocity image of the subsurface from prestack seismic recordings, once the deep-learning (DL) model is well-trained based on the self-designed geological structures and simulated recordings. However, the key problem of this approach is that it usually discards the knowledge about the forward and adjoint operators, which leads to poor reconstruction quality and generalization ability. To mitigate these problems, we have developed a deep-learning full-waveform inversion (DLFWI) approach using seismic migration images. This approach includes two key points. The first key point is that, unlike the conventional DD-DLFWI approach based on the seismic recordings in the common-source data domain, our approach utilizes the reverse time migration (RTM) images of seismic recordings in the common-source image domain as the data engine of the convolutional neural network (CNN) to reconstruct the background velocity. The second key point is that we utilize the iterative neural network architecture to reconstruct the high-resolution velocity model based on the reconstructed background velocity. Specifically, the high-resolution velocity model can be recovered by using the reconstructed velocity model, RTM image, and gradient of regularization term as the input of neural network architecture. Through synthetic experiments with various layered and fault velocity models, we have confirmed that the proposed approach can reconstruct a high-resolution velocity of the subsurface from prestack seismic recordings. Meanwhile, it outperforms the conventional DD-DLFWI approach in terms of reconstruction accuracy, antinoise, and generalization ability.

Identification of gas hydrate on bottom-simulating reflector characteristics with 2D seismic pre-stack time migration of North Bali Waters

Gas hydrate is a physical compound composed of gas molecules that are formed in a seabed layer characterised by high pressure and low temperature. It is known as one of the alternative non-conventional hydrocarbons besides petroleum and natural gas. One of the identified areas of gas hydrate stability zone is in the North Bali Waters. The North Bali Waters is part of the North East Java Basin, which has oil and gas exploration and production, both conventional and non-conventional. One method of identifying the content of gas hydrates is by looking at the appearance of the Bottom Simulating Reflector (BSR) as shown on the Pre-Stack Time Migrated seismic sections. The detection of gas hydrate zone is determined by the presence of high amplitude, reversed polarity reflection and cross-cut reflection of sedimentary layer. This study aims to determine the existence of a BSR in the waters of North Bali. The procedures for analysing the existence of Bottom Simulating Reflector in this study are pre-processing, processing, and interpretation of 2D marine seismic data. The result shows gas hydrates found with indicated Bottom Simulating Reflector on CDP 35-812 at TWT depth of 1526-1582 ms, characterised by high amplitude-reverse polarity.

Seismic Source Migration During Strombolian Eruptions Inferred by Very‐Near‐Field Broadband Seismic Network

AbstractWe analyze seismic waves excited by small Strombolian explosions to understand the source process of volcanic explosions. We deployed five broadband seismometers at only 100–300 m away from the active craters of Stromboli volcano, Italy. Moment tensor inversion of the entire seismic signals in the 0.05–0.2 Hz band locates the source at a depth of 170 and 150–200 m west/southwest of the crater where acoustic waves are excited. Contrary, the sources of seismic waves in the 0.2–0.5 and 0.5–1.0 Hz bands are excited almost at the explosion onset and are located close to the crater. We show for the first time that explosions are preceded of about 10–20 s by a small amplitude seismic phase. Semblance analysis shows that this phase is radiated from a depth of 170 m beneath the western part of the crater area. Our analysis indicates that the source moves about 50 m toward the active crater 10–20 s before the explosion occurs at the surface. At the explosion onset, the source moves back to the same location of the small preceding phase. These lateral migrations of the seismic source are estimated by moment tensor inversion and semblance analysis. We suggest that migration reflects the bending of the shallow feeding system toward northeast. Seismic waves are thus reflecting the history pressure generated by the rising of a gas‐rich pocket in the very shallow portion of a magma mush and by the following restoring force occurring after the explosion.

Reverse time migration with Gaussian beams using optimized ray tracing systems in transversely isotropic media

ABSTRACTPrestack depth migration is a key technology for imaging complex reservoirs in media with strong lateral velocity variations. Prestack migrations are broadly separated into ray‐based and wave‐equation‐based methods. Because of its efficiency and flexibility, ray‐based Kirchhoff migration is popular in the industry. However, it has difficulties in dealing with the multi‐arrivals, caustics and shadow zones. On the other hand, wave‐equation‐based methods produce images superior to that of the ray‐based methods, but they are expensive numerically, especially methods based on two‐way propagators in imaging large regions. Therefore, reverse time migration algorithms with Gaussian beams have recently been proposed to reduce the cost, as they combine the high computational efficiency of Gaussian beam migration and the high accuracy of reverse time migration. However, this method was based on the assumption that the subsurface is isotropic. As the acquired azimuth and maximum offsets increase, taking into account the influence of anisotropy on seismic migration is becoming more and more crucial. Using anisotropic ray tracing systems in terms of phase velocity, we proposed an anisotropic reverse time migration using the Gaussian beams method. We consider the influence of anisotropy on the propagation direction and calculate the amplitude of Gaussian beams with optimized correlation coefficients in dynamic ray tracing, which simplifies the calculations and improves the applicability of the proposed method. Numerical tests on anisotropic models demonstrate the efficiency and accuracy of the proposed method, which can be used to image complex structures in the presence of anisotropy in the overburden.

Technique for the imaging crystalline basement according to the DSS data

The method of deep seismic sounding (DSS), the observation systems in which are characterized by an irregular arrangement of both sources and receivers along the profile, a significant step between receivers, as well as maximum source-receiver distances exceeding several hundred kilometers, makes it possible to obtain an image of the crystalline basement using seismic migration fields of reflected/refracted waves. The main part of the existing migration methods, the use of which makes it possible to form an image of the deep structure of the study area in the dynamic characteristics of the recorded wave field, is focused on processing seismic data obtained by the method of reflected waves with multiple overlap observation systems (MOV—CDP). And, as a rule, these migration methods are designed for a smooth change in speed with depth. At the same time, at the boundary of the crystalline basement, the speed changes very sharply, which must be taken into account when processing data using migration. The proposed method for constructing an image of the crystalline basement is based on the use of finite-difference migration of the field of reflected/refracted waves, which was developed at the Institute of Geophysics named after S. I. Subbotin National Academy of Sciences of Ukraine. This migration method is designed to isolate supercritically reflected and refracted waves recorded from the basement in the far zone of the source and takes into account the full trajectory of waves passing through a two-layer medium, at the boundary of which there is a significant jump in velocity. Thus, the migration of the field of reflected/refracted waves makes it possible to obtain a correct image of the structure of the refractive layer of the crystalline basement. The article describes in detail the algorithm of the technique for constructing an image of the crystalline basement using finite-difference migration of the field of reflected/refracted waves and its difference from similar methods of migration. The advantages and disadvantages of the proposed method are shown when solving problems of regional seismic research. Explained and illustrated the features of constructing the image of violations on the border of the foundation. The effectiveness of the technique is demonstrated on a model example and real seismic data observed by the DSS method on the territory of Ukraine.

The vertical image projection method for migration stretch compensation

True amplitude inversion is often carried out without taking into account migration distortions to the wavelet. Seismic migration leaves a dip-dependent effect on the wavelet that can cause significant inaccuracies in the inverted impedances obtained from conventional inversion approaches based on 1D vertical convolutional modeling. Neglecting this effect causes misleading inversion results and leakage of dipping noise and migration artifacts from the higher frequency bands to the lower frequencies. I have observed that, despite the dip dependency of this effect, low-dip and flat events may also suffer if they are contaminated with cross-cutting noise, steep migration artifacts, and smiles. I adopt an efficient, effective, and reversible data preconditioning approach that accounts for the dip dependency of the wavelet and is applied to migrated images prior to inversion. My proposed method consists of integrating data with respect to the total wavenumber followed by the differentiation with respect to the vertical wavenumber. This process is equivalent to applying a deterministic dip-consistent preconditioning that projects the data from the total wavenumber to the vertical wavenumber axis. This preconditioning can be applied to pre- and poststack data as well as to amplitude-variation-with-offset attributes such as intercept and gradient before inversion. The vertical image projection methodology that I adopt here reduces the impact of migration artifacts such as cross-cutting noise and migration smiles and improves inverted impedances in synthetic and real data examples. In particular, I indicate that neglecting the proposed preconditioning leads to anomalously higher impedance values along the steeply dipping structures.

Near-Surface Geological Structure Seismic Wave Imaging Using the Minimum Variance Spatial Smoothing Beamforming Method

Erecting underground structures in regions with unidentified weak layers, cavities, and faults is highly dangerous and potentially disastrous. An efficient and accurate near-surface exploration method is thus of great significance for guiding construction. In near-surface detection, imaging methods suffer from artifacts that the complex structure caused and a lack of efficiency. In order to realize a rapid, accurate, robust near-surface seismic imaging, a minimum variance spatial smoothing (MVSS) beamforming method is proposed for the seismic detection and imaging of underground geological structures under a homogeneous assumption. Algorithms such as minimum variance (MV) and spatial smoothing (SS), the coherence factor (CF) matrix, and the diagonal loading (DL) methods were used to improve imaging quality. Furthermore, it was found that a signal advance correction helped improve the focusing effect in near-surface situations. The feasibility and imaging quality of MVSS beamforming are verified in cave models, layer models, and cave-layer models by numerical simulations, confirming that the MVSS beamforming method can be adapted for seismic imaging. The performance of MVSS beamforming is evaluated in the comparison with Kirchhoff migration, the DAS beamforming method, and reverse time migration. MVSS beamforming has a high computational efficiency and a higher imaging resolution. MVSS beamforming also significantly suppresses the unnecessary components in seismic signals such as S-waves, surface waves, and white noise. Moreover, compared with basic delay and sum (DAS) beamforming, MVSS beamforming has a higher vertical resolution and adaptively suppresses interferences. The results show that the MVSS beamforming imaging method might be helpful for detecting near-surface underground structures and for guiding engineering construction.

Application Of Finite Difference Eikonal Solver For Traveltime Computation In Forward Modeling And Migration

Traveltime is one of the propagating wave’s components. As the wave propagates further, the traveltime increases. It can be computed by solving wave equation of the ray path or the eikonal wave equation. Accurate method of computing traveltimes will give a significant impact on enhancing the output of seismic forward modeling and migration. In seismic forward modeling, computation of the wave’s traveltime locally by ray tracing method leads to low resolution of the resulting seismic image, especially when the subsurface is having a complex geology. However, computing the wave’s traveltime with a gridding scheme by finite difference methods able to overcomes the problem. This paper aims to discuss the ability of ray tracing and fast marching method of finite difference in obtaining a seismic image that have more similarity with its subsurface model. We illustrated the results of the traveltime computation by both methods in form of ray path projection and wavefront. We employed these methods in forward modeling and compared both resulting seismic images. Seismic migration is executed as a part of quality control (QC). We used a synthetic velocity model which based on a part of Malay Basin geology structure. Our findings shows that the seismic images produced by the application of fast marching finite difference method has better resolution than ray tracing method especially on deeper part of subsurface model.

Migration of Fluid‐Induced Seismicity Reveals the Seismogenic State of Faults

AbstractSeismic swarms, in natural contexts or induced by anthropogenic fluid injections, commonly show migrations of earthquake hypocenters. Classically, such migrations are interpreted as the diffusion of the fluid pressure. However, recent studies show that the seismic front maps the stress concentration at the tips of propagating aseismic slip that is primarily induced by the fluid pressure increase. In this case, seismic migration might depend on the hydromechanical properties that control the dynamics of aseismic slip propagation rather than the hydraulic diffusivity. Here, we use synthetic seismic catalogs obtained from the hydromechanical modeling of a slip‐weakening, permeable fault response to a fluid injection. By varying the fault permeability and the stress state, we show that the shape of the seismic front in a distance‐time plot depends on the initial fault criticality to failure, and not on the hydraulic properties. We then extrapolate this numerical result to seven observed injection‐induced earthquake swarms by showing that the shape of the migration front directly depends on the seismogenic index. The seismic front has a diffusive behavior when the aseismic slip is directly driven by the pressure increase, either at the swarm beginning or later if the faults are not initially critically stressed. On the contrary, an accelerating seismic front indicates a high‐criticality stressed fault, on which the aseismic slip runs away from the fluid pressurized area. Therefore, this study highlights that the seismic front shape may prove a useful indicator to characterize the initial stress state of faults over which the swarms occur.

Integration of Post-stack Inversion and Rock Physics to Determine Sandstone Reservoir Quality: Barakan Sub-basin case

Barakan Sub-basin is assessed as potential basin for hydrocarbon reserves in the eastern region of Indonesia because it is adjacent to Masela block giant gas field. Reservoir rocks in this sub-basin are sandstones from Middle Jurassic (Lower Flamengo Formation) until Oligocene (Adi member Formation). Main sandstone reservoir rocks are knowingly studied to have good porosity in Upper Flamengo, Kopae, Ekmai and Adi member Formations. But, there is no significant study to determine sandstone reservoir distribution that have good porosity quality. Therefore, an integrated method of inversion and rock physics study are needed to determine sandstone reservoir quality. This study uses 2D marine seismic post-stack time migration and 2 wells namely Barakan-1 and Koba-1 wells. Sensitivity analysis with cross-plot of gamma ray log versus acoustic impedance values range of 20-60 API and 9000-42000 (ft/s)*(g/cc) shows a strong correlation of good porosity sandstone to low impedance in Ekmai Formation of both wells. Model based of post-stack inversion reveals sandstone distribution in Ekmai Formation of both wells. Time structure maps of top and bottom horizons in Ekmai Formation indicates Barakan-1 well within anticline height structure and Koba-1 well are deposited in a middle of sub-littoral environment.

Intelligent Deblending of Seismic Data Based on U-Net and Transfer Learning

The blended acquisition allows multiple sources to be simulated simultaneously in a narrow time interval, which can improve the acquisition efficiency and reduce the acquisition cost tremendously. However, the overlapped information from multiple sources poses challenges for traditional seismic data migration or inversion algorithms. Thus, accurate and efficient deblending should be implemented as a pre-requisite. Traditional inversion-based deblending algorithms can provide deblended data with a high computational burden, especially for a large volume of seismic data. As a deep learning strategy can match seismic data accurately in a nonlinear way through supervised learning, we propose a U-net-based accurate deblending algorithm, which incorporates transfer learning and an iterative strategy. A set of labeled synthetic data with a blending fold of 2 are classified into the training and validation data for U-net training and validation. Field data are regarded as the test data to assess the performance of the trained U-net. To guarantee the deblending performance of the field data to some extent, parts of field data with labels are used to fine-tune the trained U-net based on transfer learning. The fine-tuning procedure is relatively fast within several minutes. To further improve the deblending performance, we incorporate an iterative strategy with the fine-tuned U-net. The deblending performance is promising in the quality and computational efficiency compared with the curvelet-thresholding-based deblending method, which demonstrates the validity of the proposed intelligent deblending method.