E-mail Alert Rss
 

物探与化探, 2023, 47(2): 384-390 doi: 10.11720/wtyht.2023.2633

方法研究·信息处理·仪器研制

基于LSCG法和波数补偿的频率域二维地震正演模拟方法

张入化,1, 张洞君1, 黄建平2, 苟其勇1, 周嘉妮3

1.中国石油西南油气田页岩气研究院,四川 成都 610051

2.中国石油大学(华东) 地球科学与技术学院,山东 青岛 266580

3.中国电建集团成都勘测设计研究院有限公司,四川 成都 610072

Frequency-domain 2D seismic forward modeling method based on the LSCG method and the wavenumber compensation

ZHANG Ru-Hua,1, ZHANG Dong-Jun1, HUANG Jian-Ping2, GOU Qi-Yong1, ZHOU Jia-Ni3

1. Shale Gas Research Institute,PetroChina Southwest Oil & Gas field Company,Chengdu 610051,China

2. School of Geosciences,China University of Petroleum(East China),Qingdao 266580,China

3. Chendu Engineering Corporation Limited,Power China,Chengdu 610072,China

第一作者: 张入化(1996-),男,硕士,助理工程师,主要研究地震资料处理和地震地质综合解释。Email:642586376@qq.com

责任编辑: 叶佩

收稿日期: 2021-12-25   修回日期: 2022-12-12  

基金资助: 西南油气田科技项目(20210304-02)

Received: 2021-12-25   Revised: 2022-12-12  

摘要

在地震勘探中,地震正演模拟是非常重要的技术。与时间域正演相比,频率域正演速度快,计算效率高。如何高效准确地完成频率域正演计算是目前该领域的一个重要问题。数值频散问题和如何提高计算效率降低求解分解阻抗内存占用量一直是频率域正演所需要解决的问题。与传统的直接法求解阻抗矩阵的频率域正演方法不同,本文采用最小二乘共轭梯度法(LSCG法)求解阻抗矩阵进行频率域正演,并提出了一种波数补偿的表达式来压制数值频散现象。经过简单模型和复杂模型的数值测试,采用最小二乘共轭梯度法(LSCG法)求解阻抗矩阵进行频率域正演能够有效降低计算时间,且采用波数补偿的频率域正演方法能够有效压制数值频散现象,提高波场模拟精度。

关键词: 地震勘探; 频率域正演; 最小二乘共轭梯度法; 数值频散; 计算效率

Abstract

The seismic forward modeling technique is critical to seismic exploration.Moreover,it shows a faster rate and higher calculation efficiency in the frequency domain than in the time domain.Presently,there is a need to complete the forward calculation in the frequency domain efficiently and accurately.The specific problems include the numerical dispersion and the high memory consumption for calculating and decomposing impedance,which should be reduced by improving the calculation efficiency.Different from the conventional direct method,this study adopted the least-squares conjugate gradient (LSCG) method used to determine the impedance matrix for the frequency-domain forward modeling and proposed an expression for wavenumber compensation to suppress the numerical dispersion.The numerical tests of simple and complex models show that the LSCG method can effectively reduce the calculation time and that the frequency-domain forward modeling method based on wavenumber compensation can effectively suppress the numerical dispersion and thus improve the precision of wave field simulation.

Keywords: seismic exploration; frequency-domain forward modeling; least-squares conjugate gradient method; numerical dispersion; calculation efficiency

PDF (3675KB) 元数据 多维度评价 相关文章 导出 EndNote| Ris| Bibtex  收藏本文

本文引用格式

张入化, 张洞君, 黄建平, 苟其勇, 周嘉妮. 基于LSCG法和波数补偿的频率域二维地震正演模拟方法[J]. 物探与化探, 2023, 47(2): 384-390 doi:10.11720/wtyht.2023.2633

ZHANG Ru-Hua, ZHANG Dong-Jun, HUANG Jian-Ping, GOU Qi-Yong, ZHOU Jia-Ni. Frequency-domain 2D seismic forward modeling method based on the LSCG method and the wavenumber compensation[J]. Geophysical and Geochemical Exploration, 2023, 47(2): 384-390 doi:10.11720/wtyht.2023.2633

0 引言

地震正演技术是研究地下介质地震响应特征的一个十分有效的手段[1-2]。近年来,各位专家学者对地震波数值模拟进行了很多研究,都取得了不少成就,但是大部分的波场模拟都是在时间域进行的。时间域计算方法是按时间片递推计算,由前一个时间片的波场值推出下一个时间片的波场值,因此前一时刻的计算误差会逐步累积到下一时刻中。当然如果计算的时间切片数目较多,计算误差将会累计,导致计算结果信噪比下降、精度太低而无法为接下来的波形反演提供研究基础。近几十年以来,研究正演的专家学者们做了大量的工作, Lysmer等[3] 最早提出频率域正演模拟方法,提高了正演模拟的计算效率;Shin等[4]进一步发展了这种方法,将频率域方法推广至地震波形反演技术; Jo等[5]提出了最优化9点差分格式,有效地减弱了数值频散现象;Min等[6]提出了一种频率域标量波动方程25点差分格式,并且计算最优化系数来压制频散现象,这种方法同样也降低了空间采样点数;Stekl等[7]将最优化9点差分方法引入到变密度声波方程和黏弹性介质的弹性波方程,提升了弹性介质的正演模拟精度;Hustedt等[8]研究了交错网格上导数算子的四阶精度有限差分格式,能够更精确地求解方程组。

频率域正演模拟主要涉及到大型稀疏矩阵的求解问题,求解的方法主要分为直接法与迭代法,直接法包括高斯消元法和LU分解法等,迭代方法有最小二乘正交分解法LSQR和最小二乘共轭梯度法LSCG等[9-10]。一般来说,直接法在进行频率域正演中求解阻抗矩阵时具有占用内存大、计算效率低的特点,因此本文采用最小二乘共轭梯度法(LSCG)求解大型稀疏矩阵,具有占用储存空间少、计算速度快等优点,我们可以利用矩阵稀疏性简化运算,适合用来求解大型稀疏矩阵。

频率域正演也存在数值频散问题,数值频散是利用有限差分数值模拟技术对时间和空间进行离散时,引入的一种影响模拟精度的现象[11-14]。数值频散主要表现为不同波数分量的地震波波形散开,形成模拟时的一种模糊假象。计算网格间距过大或者是不够精确的有限差分算法都可能会产生数值频散现象[15-17]。为了压制数值频散,普通的做法则是会提高网格密度或增加计算的网格点数,但是这两种解决办法都会导致在频率域正演求解阻抗矩阵时内存过大,计算效率降低。

本文在计算频率域地震正演中,利用最小二乘共轭梯度法对稀疏矩阵进行求解,并从波动方程本身出发,从理论上分析数值频散产生的原因,对高波数进行补偿得到一种声波方程。最后通过简单地震模型和复杂地震模型测试,对计算效率和数值频散压制效果进行了验证。

1 基本原理

1.1 LSCG法求解波场

本文采用迭代方法中的最小二乘共轭梯度法(LSCG)来求解阻抗矩阵,声波方程一般采用九点差分格式(图1)。

图1

图1   九点差分格式

Fig.1   Nine-point difference format


本文将波场空间划分横向上为Nx,纵向上为Nz的网格空间。设n=Nx×Nz,

M(v,ω)U(ω)=F(ω)

式中:M是复阻抗系数矩阵,大小为n×n;U(ω)是频率域中的离散波场;F(ω)是震源项。本文将复数矩阵转化为实系数分块矩阵,那么可以表示成:

(MRe+iMIm)(URe+iUIm)=FRe+iFIm

式中,MReMIm分别是阻抗系数矩阵的实部和虚部,UReUIm分别是波场矩阵的实部和虚部,FReFIm分别是震源矩阵的实部和虚部。将上式展开并写成分块矩阵的形式,如下:

MRe-MImMIm MReUReUIm=FReFIm

本文将复阻抗系数矩阵M进行排列,只对非零元素进行存储,然后采用LSCG方法进行求解。求解形式可简单写成:

An×nun×1=bn×1

假设u0是要求解的波场的初始值,β0是已知的震源信息的二范数。最小二乘共轭梯度法的求解过程如下:

u0=0,r0=b0-Au0,i=0Whileri0i=i+1ifi=1p0=r0elseβi-2=rTi-1ri-1/rTi-2ri-2Pi-1=ri-1+βi-2Pi-2endαi-1=rTi-1ri-1/PTi-1APi-1ui=ui-1+αi-1Pi-1ri=ri-1-αi-1APi-2endu=ui+1

在上述求解过程中,ui+1为最后输出的波场虚实分解后的列向量,将实部和虚部结合就可以得到波场解。与常规正演方法相比,频率域下最小二乘共轭梯度波动方程正演,对波场虚部和实部分别进行存储,并同时采用了稀疏矩阵存储的方法,因此计算速度较快,占用内存相对较小。

图2为基于最小二乘共轭梯度法求解的频率域正演方法流程。

图2

图2   频率域正演流程

Fig.2   Frequency domain forward flow chart


1.2 波数补偿

数值频散主要是因为高波数分量滞后较多而造成的。本文构思了一种改进方程对波数进行补偿。c为补偿算子,下式为时间域方程:

(v22-22t+c2t)u=0

其中,c=12σv,y=c2,y的空间傅立叶变换是波数k的增长函数。σ是稳定因子,本文经过多轮测试,其稳定因子具体大小对结果并没有明显的影响,一般取接近于0的正值,本文设置的稳定因子为e-6

方程(5)可以写成以下形式:

(2yvσ-22t+2yt)u=0

我们可以定义一个中转波场z:

z=ue-yt

再将式(7)和式(6)写成:

u=zeyt,[(2vσy+1)y2-22t]u=0 ,

式中:v是声波速度;u是时间域波场。

最后将其转化为频率域下的改进方程:

ω2v2U(x,z,ω)+(1+iωσv)2U(x,z,ω)=0

式(8)与常规的声波波动方程相似,我们可以将定义的中转波场看作是常规波动方程的解,而原来的时间域波场在式(9)中可以看作是关于波数k的指数增长模式。那么本文所的改进方程则是在常规声波方程的基础上对波数的一个补偿,对高波数的补偿更为强烈。数值频散现象本就是由高波数分量滞后较多造成的。本文的波数补偿方法能够对高波数分量进行一定程度的补偿,进而压制数值频散,提高正演精度。

2 数值模拟

简单模型采用洼陷模型(图3),模型大小为201×201,网格间距为10 m,正演所采用雷克子波主频为30 Hz。简单模型在进行数值测试中,采用PML边界进行边界吸收,设置的PML边界厚度是30。图3为洼陷模型。对于简单洼陷模型,本文采用LSCG+波数补偿方法的九点差分频率域正演模拟和常规频率域正演方法进行对比。图4分别是频率20 Hz(图4a)和36 Hz(图4b)的频率切片。表1是洼陷模型正演中两种不同求解阻抗矩阵的正演方法所需计算时间和占用内存的对比分析。我们可以发现,最小二乘共轭梯度法计算的时间以及内存占用均少于其他方法。

图3

图3   洼陷模型

Fig.3   Sag model


图4

图4   洼陷模型正演中频率20 Hz(a)和36 Hz(b)频率切片

Fig.4   Frequency slices of 20 Hz(a) and 36 Hz(b) in the forward modeling of sag model


表1   洼陷模型正演计算效率对比

Table 1  Comparison of forward calculation efficiency of sag model

计算效率直接高斯
消元法
最小二乘共
轭梯度LSCG
不完全
LU分解
最小二乘
正交法
计算时间/s95.672.384.774.5
占用内存/GB1.50.31.40.4
单次求解方程
平均耗时/s
1.351.031.191.06

新窗口打开| 下载CSV


图5是洼陷模型的单炮记录,可以发现图5b有较为严重的数值频散现象,而图5a的数值频散现象被压制得较好。

图5

图5   洼陷模型正演单炮记录

a—采用波数补偿;b—未采用波数补偿

Fig.5   Single shot record for forward modeling of sag model

a—using wavenumber compensation;b—no wavenumber compensation


复杂模型采用加拿大洛基山脉模型(图6),模型大小260×394,网格间距10 m,雷克子波主频30 Hz。复杂模型在进行数值测试中,采用的PML边界进行边界吸收,设置的PML边界厚度是30。本文采用LSCG+波数补偿方法的九点差分频率域正演模拟。

图6

图6   加拿大洛基山脉模型

Fig.6   Canadian rockies model


图7分别为频率20 Hz(图7a)和36 Hz(图7b)的频率切片。频率切片提供了较为丰富的波场信息,可以发现在波场边界处的吸收效果比较好。取1.2 s处的时间域波场快照(图8),分别采用LSCG+波数补偿方法正演以及普通9点差分频率域正演进行对比分析。在图8的波场快照中,我们同时可以观察到有效波、数值频散以及边界处的轻微反射,未采用波数补偿的正演方法具有一定的数值频散现象,而采用波数补偿的正演方法能很好的压制数值频散现象。为了验证文中方法的抗频散能力,还在35 Hz震源频率条件下对不同方法进行了验证,如图9所示。在35 Hz的震源频率下,采用波数补偿的正演方法仍然能很好地压制数值频散现象,但是正演模拟的精度并不能通过提高震源主频而提高。表2是采用LSCG方法、直接高斯消元法、不完全LU分解法、最小二乘正交法正演方法的效率对比,可以发现,采用LSCG方法进行频率域正演模拟的所需时间比另外3种方法的频率域正演所需时间更少,效率更高。

图7

图7   加拿大洛基山脉模型正演中频率20 Hz(a)和36 Hz(b)频率切片

Fig.7   Frequency slices of 20 Hz(a) and 36 Hz(b) in the forward modeling of Canadian rockies model


图8

图8   1.2 s处的波场快照(25 Hz)

a—采用LSCG+波数补偿;b—未采用波数补偿

Fig.8   Snapshot of the wave field at 1.2 s(25 Hz)

a—using wavenumber compensation;b—no wavenumber compensation


图9

图9   1.2 s处的波场快照(35 Hz)

a—采用LSCG+波数补偿;b—未采用波数补偿

Fig.9   Snapshot of the wave field at 1.2 s(35 Hz)

a—using wavenumber compensation;b—no wavenumber compensation


表2   复杂模型正演中4种不同求解阻抗矩阵的正演方法计算效率

Table 2  Computational time required for four different forward modeling methods for solving impedance matrix in complex model forward modeling

计算效率直接高斯
消元法
最小二乘共
轭梯度LSCG
不完全
LU分解法
最小二乘
正交法
计算时间/s153.3127.2148.3130.3
占用内存/GB2.30.562.10.7
单次求解方程
平均耗时/s
1.911.591.851.63

新窗口打开| 下载CSV


图10为加拿大洛基山脉模型单炮记录。数值频散现象会造成记录的干扰信息严重增多,数值频散主要表现为高频假象。图10a是采用过波数补偿方法的正演模拟,图10b是未采用波数补偿方法的正演模拟。可以发现图10b有较为严重的数值频散现象,而图10a的数值频散现象被压制得较好,且波场较为清晰。图11是分别对采用波数补偿和未采用波数补偿的地震正演单炮记录抽取第100道进行对比分析。我们可以发现未采用波数补偿的地震单炮单道记录中存在一定的噪声,而采用波数补偿地震单炮单道记录具有较少的噪声成分,这种噪声成分主要是数值频散导致的。

图10

图10   加拿大洛基山脉模型正演单炮记录

a—采用波数补偿;b—未采用波数补偿

Fig.10   The Canadian rockies model forwards a single shot record

a—using wavenumber compensation;b—no wavenumber compensation


图11

图11   正演模拟第100道单道对比分析

a—未采用波数补偿;b—采用波数补偿

Fig.11   Comparison and analysis of the 100th single-channel forward simulation

a—using wavenumber compensation;b—no wavenumber compensation


为了进一步验证算法的准确性,将本文方法和解析解方法进行模拟对比,采用速度为3 500 m/s,网格大小为200×200,网格间距为10 m的均匀介质进行模拟,主频为40 Hz,采样时间为1 s。将本文方法计算得到的距离原点为130 m的接收点信号与解析解方法得到的信号进行对比。由图12可见,本文方法接收信号与解析解信号存在一定偏差,但整体偏差较小。

图12

图12   解析解信号与本文方法接收信号对比

Fig.12   Comparison of analytical solution signal and received signal by the method in this paper


3 结论

本文针对频率域正演模拟中数值频散和计算效率问题进行了研究。采用最小二乘共轭梯度法(LSCG)对阻抗矩阵求解来进行频率域正演,并提出一种补偿方程对高波数进行补偿来进一步压制数值频散。经过洼陷模型和加拿大洛基山脉模型的数值测试,可以得到以下结论:

1)在频率域正演中,相对于常规高斯消元法求解阻抗矩阵,LSCG方法具有更高的计算效率,并且占用更少的计算内存。

2)本文提出的高波数补偿方法能够在一定程度上压制数值频散现象。

参考文献

陈可洋.

各向异性弹性介质方向行波波场分离正演数值模拟

[J]. 岩性油气藏, 2014, 26(5):91-96.

[本文引用: 1]

Cheng K Y.

Numerical simulation of forward modeling of wavefield separation of directional traveling waves in anisotropic elastic media

[J]. Lithologic Reservoirs, 2014, 26(5):91-96.

[本文引用: 1]

苏巍. 河流相砂岩储层地震响应特征与识别技术研究[D]. 长春: 吉林大学, 2007.

[本文引用: 1]

Su W. A study on seismic response characteristic and identification technology of fluvial sandstone reservoir[D]. Changchun: Jilin University, 2007.

[本文引用: 1]

Lysmer J, Drake L A.

A finite element method for seismology

[J]. Methods in Computational Physics:Advances in Research and Applications, 1972, 11:181-216.

[本文引用: 1]

Shin C S. Nonlinear elastic wave Inversion by Blocky Parameterization[D]. Tulsa: University of Tulsa, 1988.

[本文引用: 1]

Jo C, Shin C, Suh J.

An optimal 9-point finite-difference frequency space 2-D scalar wave extrapolator

[J]. Geophysics, 1996, 61(2):529-537.

DOI:10.1190/1.1443979      URL     [本文引用: 1]

In this study, a new finite‐difference technique is designed to reduce the number of grid points needed in frequency‐space domain modeling. The new algorithm uses optimal nine‐point operators for the approximation of the Laplacian and the mass acceleration terms. The coefficients can be found by using the steepest descent method so that the best normalized phase curves can be obtained. This method reduces the number of grid points per wavelength to 4 or less, with consequent reductions of computer memory and CPU time that are factors of tens less than those involved in the conventional second‐order approximation formula when a band type solver is used on a scalar machine.

Min J, Shin C, Kwon B, et al.

Improved frequency-domain elastic wave modeing using weighted-averaging difference operators

[J]. Geophysics, 2000, 65(3):884-895.

DOI:10.1190/1.1444785      URL     [本文引用: 1]

We develop a new finite‐difference scheme that reduces the number of grid points per wavelength required in frequency‐domain elastic modeling. Our approach computes weighted averages of the spatial second‐order derivative and the mass acceleration terms using a 25-point computational stencil. By determining the weighting coefficients to minimize numerical dispersion and numerical anisotropy, we reduce the number of grid points to 3.3 per shear wavelength, with a resulting error in velocities smaller than 1%. Our choice of grid points reduces the computer memory needed to store the complex impedance matrix to 4% of that for a conventional second‐order scheme and to 54% of that for a combined second‐order scheme. The 25-point weighted averaging scheme of this paper makes it possible to accurately simulate realistic models. Numerical examples show that this technique can achieve the same accurate solutions with fewer grid points than those from previous frequency‐domain second‐order schemes. Our technique can be extended directly to 3-D elastic modeling; the computational efficiency will be even greater than that realized for 2-D models.

Stekl I, Pratt K G.

Frequency-domain finite accurate difference visco-elastic modeing by using rotated operators

[J]. Geophysics, 1998, 63(5):1779-1794.

DOI:10.1190/1.1444472      URL     [本文引用: 1]

The viscoelastic wave equation is an integro‐differential equation that requires special methods when using time‐domain numerical finite‐difference methods. In the frequency domain, the integral terms are easily represented by complex valued elastic media properties. There are further significant advantages to using the frequency domain if the forward or the inverse problem requires modeling or inverting a large number of prestack source gathers. Numerical modeling is expensive for seismic data because of the large number of wavelenghths typically separating sources from receivers, which results in a need for a large number of grid points. A major obstacle to using frequency‐domain methods is the consequent storage requirements. To reduce these, we maximize the accuracy and simultaneously minimize the spatial extent of the numerical operators. We achieve this by extending earlier published methods introduced for the viscoacoustic case to the viscoelastic case. This requires the formulation of two new numerical operators: a differencing operator in a rotated coordinate frame and a lumped mass term. The new operators are combined with ordinary second‐ order, finite‐difference operators in an optimal manner to minimize numerical errors without increasing the size of the numerical operator. For a fixed number of grid points, the resulting second‐ order differencing scheme is no more expensive than an ordinary second‐order differencing scheme, but a numerical dispersion analysis shows that the number of grid points required per smallest wavelength is reduced from approximately 15 to approximately 4. The new scheme is also capable of handling embedded fluid layers without instability. We demonstrate that no further improvement in performance can be achieved using higher order spatial operators because of the associated computational overheads associated with the larger differencing operators. The new viscoelastic modeling scheme is used to study a crosshole data set in which the exact nature of the seismic coda is unclear. The results of the modeling study indicate this coda is likely related to the generation of mode‐converted shear waves within the complicated, finely layered sediments at the site.

Hustedt B, Opert S, Virieux J.

Mixed-grid and staggered grid finite-difference methods for frequency-domain acoustic wave modeling

[J]. Geophysics, 2004, 157(3):1269-1296.

[本文引用: 1]

Paige C C, Saunders M A.

LSQR:An algorithm for sparse liner equation and sparse least squares

[J]. ACM Transaction Mathematical Software, 1982, 8(1):43-71.

DOI:10.1145/355984.355989      URL     [本文引用: 1]

Bjorck A, Duff I S.

A direct method for the solution of sparse liner least squares problems

[J]. Linear Algebra and its Applications, 1980, 34(1):43-67.

DOI:10.1016/0024-3795(80)90158-5      URL     [本文引用: 1]

张京思, 揣媛媛, 边立恩.

正演模拟技术在渤海油田 X 井区砂体连通性研究中的应用

[J]. 岩性油气藏, 2016, 28(3):127-132.

[本文引用: 1]

Zhang J S, Tuan Y Y, Bian L E.

Application of forward modeling to study of sand body connectivityin X well field of Bohai Oilfield

[J]. Lithologic Reservoirs, 2016, 28(3):127-132.

[本文引用: 1]

冯德山, 王向宇.

基于卷积完全匹配层的旋转交错网格高阶差分法模拟弹性波传播

[J]. 物探与化探, 2018, 42(4):766-776.

[本文引用: 1]

Feng D S, Wang X Y.

Elastic wave propagation simulation in anisotropic media and random media using high-order difference method of rotation staggered grids based on convolutional perfectly matched layer

[J]. Geophysical and Geochemical Exploration, 2018, 42( 4) :766-776.

[本文引用: 1]

袁茂林, 蒋福友, 杨鸿飞, .

高斯束线性正演模拟方法研究

[J]. 物探与化探, 2017, 41(5):881-889.

[本文引用: 1]

Yuan M L, Jiang F Y, Yang H F, et al.

Gaussian-beam linear forward modeling

[J]. Geophysical and Geochemical Exploration, 2017, 41(5):881-889.

[本文引用: 1]

李胜军, 刘伟方, 高建虎.

正演模拟技术在碳酸盐岩溶洞响应特征研究中的应用

[J]. 岩性油气藏, 2011, 23(4):106-109.

[本文引用: 1]

Li S J, Liu W F, Gao J H.

Application of forward modeling to research of carbonate cave response

[J]. Lithologic Reservoirs, 2011, 23(4):106-109.

[本文引用: 1]

王光文, 王海燕, 李洪强, .

地震正演技术在深反射地震剖面探测中的应用

[J]. 物探与化探, 2021, 45(4):970-980.

[本文引用: 1]

Wang G W, Wang H Y, Li H Q, et al.

Application of seismic forward simulation technology in deep reflection seismic profile detection

[J]. Geophysical and Geochemical Exploration, 2021, 45( 4) :970-980.

[本文引用: 1]

董良国, 李培明.

地震波传播数值模拟中的频散问题

[J]. 天然气工业, 2004, 24(6):53-56.

[本文引用: 1]

Dong L G, Li P M.

Dispersion problem in numerical simulation of seismic wave propagation

[J]. Natural Gas Industry, 2004, 24(6):53-56.

[本文引用: 1]

曹书红, 陈景波.

声波方程频率域高精度正演的17点格式及数值实现

[J]. 地球物理学报, 2012, 55(10):3440-3449.

[本文引用: 1]

Cao S H, Chen J B.

A 17-point Scheme and its numerical implementation for high-accuracy modeling of frequency-domain acoustic equation

[J]. Chinese Journal of Geophysics, 2012, 55(10):3440-3449.

[本文引用: 1]

/

京ICP备05055290号-3
版权所有 © 2021《物探与化探》编辑部
通讯地址:北京市学院路29号航遥中心 邮编:100083
电话:010-62060192;62060193 E-mail:whtbjb@sina.com