Figure 4 - uploaded by Hongwei Liu
Content may be subject to copyright.
The flowcharts of reverse time migration of different algorithms. (a) The traditional flowchart of GPU/CPU collaborative computing RTM which needs to store the wave propagation history; (b) the flowchart of GPU/CPU collaborative computing RTM using random boundary condition.
Source publication
Prestack reverse time migration (RTM) is a very useful tool for seismic imaging but has mainly three bottlenecks: highly intensive computation cost, low‐frequency band imaging noise and massive memory demand. Traditionally, PC‐clusters with thousands of computation nodes are used to perform RTM but it is too expensive for small companies and oilfie...
Contexts in source publication
Context 1
... the one hand the wavefront has become random noise that would not form a continuous event during the imaging and on the other hand, the boundary values are used to reform the useful wavefield during back-propagation. Figure 4(a) is the original flowchart of GPU/CPU collaborative computing RTM; Fig 4(b) is the flowchart of GPU/CPU collaborative computing RTM using the pseudo random boundary condition. There are several advantages to using the random boundary condition in RTM: ...
Context 2
... the one hand the wavefront has become random noise that would not form a continuous event during the imaging and on the other hand, the boundary values are used to reform the useful wavefield during back-propagation. Figure 4(a) is the original flowchart of GPU/CPU collaborative computing RTM; Fig 4(b) is the flowchart of GPU/CPU collaborative computing RTM using the pseudo random boundary condition. There are several advantages to using the random boundary condition in RTM: ...
Similar publications
- [...]
Pre-stack reverse time migration (RTM) is a very useful tool for seismic imaging. But it has not been widely used because of the highly intensive computation cost, imaging noise and massy memory demand. In this paper, we illustrate the implementation process of RTM and analyze the stability condition and dispersion relation of finite difference (FD...
Pre-stack reverse time migration (RTM) is a very useful tool for seismic imaging, however, it has some problems such as highly intensive computation cost, imaging noise, and massy memory demand, which the researchers in the field have been striving to solve. The problem of intensive computation cost could be solved by GPU/CPU collaborative computin...
Pre-stack reverse time migration (RTM) is a very useful tool for seismic imaging. It has, however, some problems such as highly intensive computation cost, low-frequency imaging noise and massy memory demand. The problem of time consuming for calculation can be solved by GPU/CPU collaborative computation. This paper focuses on suppressing noise and...
Citations
... Nevertheless, the storage demand issue remained in the RTM-based GPU implementations presented by the earlier research. For this, Liu et al. [2012] implemented the RTM with RBC to diminish the storage demand need in migration algorithms showing that such a strategy is beneficial for GPU implementations. The GPU computational implementation with the RBC technique was coded in CUDA and tested only for 2-D RTM applications. ...
Seismic imaging is a computationally demanding and data-intensive activity in the oil and gas industry. Reverse Time Migration (RTM) used in seismic applications needs to store the forward-propagated wavefield (or source wavefield) on disk. Aiming to mitigate the storage demand, we develop an RTM that implements the source wavefield reconstruction by introducing a new wave equation to the problem. We adjust the initial and boundary conditions to take advantage of the properties of random boundary conditions (RBC). The RBC does not suppress unwanted waves coming from the artificial boundary enabling the full wavefield recovery. Besides, it explores low correlations with non-coherent signals due to the random velocities in the boundary. We also develop compiler-guided implementations on a vector processor for seismic modeling and RTM, essential for Least-square Migration, Full Waveform Inversion, and Uncertainty Quantification applications. We test the seismic modeling and RTM on the 2-D Marmousi benchmark and 3-D HPC4E Seismic Test Suite. The numerical experiments show that the RTM which implements the wavefield reconstruction presents the best results in terms of execution time and hard disk demand. Lastly, the vector processor implementation is the one that requires fewer code modifications compared to the optimized baseline versions of the seismic modeling and RTM and GPU implementations, particularly for large 3D grids.
... The application of NVIDIA GPUs in seismic image processing has significantly improved computing efficiency, for instance, in time migration and RTM and waveform inversion. [15][16][17], especially RTM, because it has the highest calculation accuracy in the seismic migration but is also the most time-consuming. Therefore, when GPUs appeared, it was used to accelerate RTM computation [18] by includingthe useof random boundary instead of storing wavefields to save GPU global memory [19], as well as some GPU computing strategies for the calculation of specific form of RTM tilted transversely isotropic TTI RTM and Q-RTM [20,21]. ...
As an important method for seismic data processing, reverse time migration (RTM) has high precision but involves high-intensity calculations. The calculation an RTM surface offset (shot–receiver distance) domain gathers provides intermediary data for an iterative calculation of migration and its velocity building. How to generate such data efficiently is of great significance to the industrial application of RTM. We propose a method for the calculation of surface offset gathers (SOGs) based on attribute migration, wherein, using migration calculations performed twice, the attribute profile of the surface offsets can be obtained, thus the image results can be sorted into offset gathers. Aiming at the problem of high-intensity computations required for RTM, we put forth a multi-graphic processing unit (GPU) calculative strategy, i.e., by distributing image computational domains to different GPUs for computation and by using the method of multi-stream calculations to conceal data transmission between GPUs. Ultimately, the computing original efficiency was higher relative to a single GPU, and more GPUs were used linearly. The test with a model showed that the attributive migration methods can correctly output SOGs, while the GPU parallel computation can effectively improve the computing efficiency. Therefore, it is of practical importance for this method to be expanded and applied in industries.
... In recent years, Graphic Processing Unit (GPU) parallel computing has gained more and more popularity in many scientific fields, such as physics, geophysics, and computer science (Preis, 2011;Panetta et al., 2009). A number of geophysical problems have been solved using GPUs, such as wavefield simulation (Michéa and Komatisch, 2010), reverse time migration (Micikevicius, 2009;Abdelkhalek et al., 2009;Liu et al., 2012) and full waveform inversion (Shin et al., 2014;Yang et al., 2015;Liu et al., 2015). In this paper, we carry out parallel computing using a NVIDIA ® GPU card (Tesla C2075) to accelerate the numerical simulation of seismic wave propagation. ...
Seismic wavefield simulation in the presence of surface topography provides important information for characterizing seismic wave propagation. Based on the boundary-conforming grid method, we first transform both elastic wave equations in second-order formulation and free surface boundary condition in first order system from Cartesian coordinates to curvilinear coordinates. Then, the convolutional perfectly matched layer (CPML) boundary condition is applied to absorb the outgoing seismic waves at the edges of the truncated model. The test results (e.g., wavefield snapshots and seismograms) show that our numerical algorithms can effectively simulate seismic wave propagation in a model with rough topography, and CPML is more efficient than perfectly matched layer (PML) boundary condition in suppressing artificial reflections. In addition, the finite-difference algorithms on a single Graphic Processing Unit (GPU) are used to accelerate seismic numerical modeling in both elastic isotropic and anisotropic media. Compared with the conventional CPU version, the GPU implementation greatly reduces the computational cost.
... In addition, the development tools known as compute unified device architecture (CUDA) provide an easily manageable way to edit and compile code for execution on GPUs. GPU implementations with CUDA have been widely used in seismic migration imaging and wavefield modelling (Zhang et al., 2009a,b;Rached A. et al., 2012;Liu et al., 2012;Liu et al., 2013;Weiss and Shragge, 2013;Venstad, 2016). Therefore, in this paper we demonstrate the GPU-based focal beam analysis algorithm for 3D acquisition geometries in complex media. ...
Seismic acquisition geometries have a significant influence on the quality of seismic data in the oil and gas exploration process. Therefore, prior analyses are beneficial to the design of acquisition geometries before implementation of seismic acquisition. The focal beam analysis method can provide quantitative insights into the combined influence of acquisition geometries and subsurface structures. This approach involves a large calculation burden concerning 3D wavefield extrapolation in the case of complex media, thus inhibiting the practical application of focal beam analysis in complex media when using regular CPUs. Therefore, using a graphics processing unit (GPU) to accelerate focal beam analysis becomes imperative. We have developed a fast parallel algorithm to speed up the focal beam analysis for 3D acquisition geometries in complex media on GPUs. Three-dimensional numerical examples show that the GPU-based focal beam analysis runs about 17 times faster than a serial CPU-based one. We also demonstrate the validity and scalability of the proposed approach with numerical examples. The boost in performance afforded by the GPU architecture allows us to analyse 3D acquisition geometries in complex media with less time and at lower cost of hardware.
... The RTM is actively being used in the oil and gas industries where several techniques have been developed to alleviate these bottlenecks. To speed up computations, oil and gas industries use parallel computing techniques [27][28][29][30][31]. To alleviate the memory bottleneck, algorithms have been developed [32][33][34][35][36]. ...
... To obtain the time history of motion of one particle of the medium, we consider the wave equation corresponding to anti-plane shear. Assuming that the waves propagate in the x-z plane and are polarized in the y-direction, the only nonzero displacement component is u y and the wave equation is [31]: ...
The emergence of ultrasonic dry point contact (DPC) transducers that emit horizontal shear waves has enabled efficient collection of high-quality data in the context of a nondestructive evaluation of concrete structures. This offers an opportunity to improve the quality of evaluation by adapting advanced imaging techniques. Reverse time migration (RTM) is a simulation-based reconstruction technique that offers advantages over conventional methods, such as the synthetic aperture focusing technique. RTM is capable of imaging boundaries and interfaces with steep slopes and the bottom boundaries of inclusions and defects. However, this imaging technique requires a massive amount of memory and its computation cost is high. In this study, both bottlenecks of the RTM are resolved when shear transducers are used for data acquisition. An analytical approach was developed to obtain the source and receiver wavefields needed for imaging using reverse time migration. It is shown that the proposed analytical approach not only eliminates the high memory demand, but also drastically reduces the computation time from days to minutes.
... Currently, the parallel graphic processing unit (GPU) becomes prevailing in non-graphical applications for its powerful computational capacity and high computational efficiency, thus attracting increasing interests in seismic exploration (e.g., Krakiwsky et al., 2004;Inman et al., 2007;Price et al., 2007;Micikevicius, 2009;Mich ea and Komatitsch, 2010;Liu et al., 2012a;Kim et al., 2013;Weiss and Shragge, 2013;Shin et al., 2014). The RTM implementation on GPU has proven to be able to greatly improve the computational efficiency (e.g., Abdelkhalek et al., 2009;Liu et al., 2012b;Gao et al., 2014;Yang et al., 2014;Shi and Wang, 2015;Li et al., 2017). ...
... The second is the ABC. Reflections from limited model boundaries would degrade final images, though the random boundary condition (Clapp, 2009) could reduce the memory burden and improve the efficiency on GPU (e.g., Liu et al., 2012bLiu et al., , 2013Yang et al., 2014), it still introduces strong artificial reflections from model boundaries. Therefore, other ABCs are in demand. ...
... NVIDIA's GPU contain an architecture known as compute unified device architecture (CUDA), its language is an extension to the C programming language. The parallel algorithm on GPU with CUDA could greatly improve the computational efficiency of RTM (e.g., Abdelkhalek et al., 2009;Liu et al., 2012b;Gao et al., 2014;Shi and Wang, 2015). In addition, the POSIX thread, enabling parallel computing on classical central processing unit (CPU), also could save computational time and is easy to be utilized (Wagner and Towsley, 1995;Liu et al., 2011b). ...
Reverse-time migration (RTM) is a powerful tool for imaging geologically complex structures such as steep-dip and subsalt. However, its implementation is quite computationally expensive. Recently, as a low-cost solution, the graphic processing unit (GPU) was introduced to improve the efficiency of RTM. In the paper, we develop three ameliorative strategies to implement RTM on GPU card. First, given the high accuracy and efficiency of the adaptive optimal finite-difference (FD) method based on least squares (LS) on central processing unit (CPU), we study the optimal LS-based FD method on GPU. Second, we develop the CPU-based hybrid absorbing boundary condition (ABC) to the GPU-based one by addressing two issues of the former when introduced to GPU card: time-consuming and chaotic threads. Third, for large-scale data, the combinatorial strategy for optimal checkpointing and efficient boundary storage is introduced for the trade-off between memory and recomputation. To save the time of communication between host and disk, the portable operating system interface (POSIX) thread is utilized to create the other CPU core at the checkpoints. Applications of the three strategies on GPU with the compute unified device architecture (CUDA) programming language in RTM demonstrate their efficiency and validity.
... In 2006, NVIDIA released a Compute Unified Device Architecture (CUDA), mathematic works can run in parallel using GPU. After that, high order finite-difference stencil solving problem can be implemented with GPU efficiently (Micikevicius, 2009) and RTM implementation is introduced using GPU (Liu et al., 2012). RTM based on MPI and GPU separately has been implemented in different homogenous and inhomogeneous platforms (Suh et al., 2010;Liu et al., 2012). ...
... After that, high order finite-difference stencil solving problem can be implemented with GPU efficiently (Micikevicius, 2009) and RTM implementation is introduced using GPU (Liu et al., 2012). RTM based on MPI and GPU separately has been implemented in different homogenous and inhomogeneous platforms (Suh et al., 2010;Liu et al., 2012). Combining them together, MPI & CUDA hybrid accelerating method can be used in the amplitude reserved Q compensation prestack time migration to greatly reduce the calculation time (Han and Sun, 2015). ...
... Pedersen et al. (2010) derived phase slowness expressions for P-and SV-waves that are used in a one-way wave-equation migration scheme in VTI media. In addition, many studies about anisotropic reverse time migration (RTM) have been proposed (Du et al. 2007;Fletcher et al. 2009;Fowler et al. 2010;Duveneck and Bakker 2011;Liu et al. 2012;Zhan et al. 2012). ...
An approach for extracting angle-domain common-image gathers (ADCIGs) from anisotropic Gaussian beam prestack depth migration (GB-PSDM) is presented in this paper. The propagation angle is calculated in the process of migration using the real-value traveltime information of Gaussian beam. Based on the above, we further investigate the effects of anisotropy on GB-PSDM, where the corresponding ADCIGs are extracted to assess the quality of migration images. The test results of the VTI syncline model and the TTI thrust sheet model show that anisotropic parameters ε, δ, and tilt angle 𝜃, have a great influence on the accuracy of the migrated image in anisotropic media, and ignoring any one of them will cause obvious imaging errors. The anisotropic GB-PSDM with the true anisotropic parameters can obtain more accurate seismic images of subsurface structures in anisotropic media.
... Moreover, it is worth mentioning that synthetic datasets, as Marmousi and SEG/EAGE models, are commonly used for evaluating geophysical solutions, techniques, and algorithms, because they faithfully represent geological features found in real datasets [12,[43][44][45][46]. ...
The Pre-stack Kirchhoff Time Migration (PKTM) is a central process in petroleum exploration. As PKTM is computationally intensive, many works have proposed the use of accelerators like GPU and FPGA to improve its execution time. On the other hand, although many off-the-shelf processors are endowed with a set of SIMD vector instructions, few papers tackle the problem considering vectorization and all of them consider that compilers can successfully vectorize the code. In this paper, we show that programming PKTM by using SIMD vector instructions manually is more efficient than the automatically and semi-automatically vectorized codes, provided by a hardware specific compiler and library. Experiments considering both real and synthetic datasets showed that our solution is more than four times faster than the traditional code. It also outperformed automatically vectorized codes in all tests. We believe that the proposed strategy can be used together with the other ones to accelerate seismic migration methods in general without new investments in hardware. Copyright
... Naturally, RTM is the first choice to improve the imaging quality of the salt dome structures in Pre-Caspian Basin. However, the verification of RTM is using the salt dome model of the Gulf of Mexico deep-water oil field (Liu et al. 2009;Zhang et al. 2007a, b;Liu et al. 2012). It is necessary to analyze and compare the imaging capabilities of RTM and WPDM based on the actual regional geologic characteristics, and then give some suggestions to apply the migration method to the real seismic data processing. ...
The Pre-Caspian Basin is one of the most prolific petroliferous basins in Kazakhstan. However, the hydrocarbon reservoir is always located in salt dome and pre-salt structures. The salt-related structure is so complex that it is difficult to obtain the satisfied imaging results. In order to improve the imaging precision and provide authentic amplitude information to reservoir prediction, we should analyze the imaging and amplitude preserving capability between different migration methods. According to actual geologic and reservoir characteristics, we design a typical complex salt-related structure model of the basin. Based on the forward modelling seismic record, we compare the one-way wave-equation pre-stack depth migration (WPDM) and pre-stack reverse-time depth migration (RTM). The results show that RTM can greatly improve the imaging quality of salt dome, but its improvement for the pre-salt structures is small compared with WPDM. Both methods are equally capable of preserving seismic reflection characteristics that affected mainly by the lateral variation of overlying strata.
