基于GPU和CPU协同并行的三维各向异性介质地震波场正演模拟
|
摘要
莺歌海盆地中深部地层具有多套超低速层、异常高压、垂向裂隙发育等特点,使得介质具有各向异性,地震波场特征复杂,正演模拟是研究波场特征和观测系统优化的主要手段之一,而海上震源子频带宽、主频高,要求正演模拟网格剖分小,导致计算量大。为此,发展了基于GPU和CPU协同并行的海上三维各向异性介质正演模拟方法,通过将模型分割并分配到不同进程上和任意选择并行计算的方向和每个方向上并行计算的进程个数,不仅可以减小每个进程上内存消耗,而且减少计算时间。简单3D模型的正演模拟验证了该方法可极大地提高计算效率,复杂构造各向异性介质模型中的模拟炮集记录的偏移成像结果验证了方法的可靠性,可适用于任意各向异性介质地震波场正演模拟。
There are many ultra-low-speed layers, abnormal high pressure and vertical fracture in deep strata at Yinggehai Basin, which makes anisotropic and structure complex. The characteristics of seismic wavefield are complicated. Forward modelling is one of the main methods to study wavefield characteristics and observation system optimization. However, broad band and high frequency of the source wavelet in marine as well as the small grid size of the model result in a large amount of calculation. To better the calculation, the forward modelling of seismic wavefield in 3 D anisotropic media is developed based on GPU and CPU collaboration parallel in this paper. The 3 D model is divided and assigned to different processes, and the direction of parallel computing and the number of processes in each direction is arbitrarily selected, which can not only reduce the memory consumption in each process, but also reduce the calculation time. The forward modelling of seismic wavefield in a simple 3 D model verifies the method and can greatly improve the computational efficiency. The migration results of the synthetic common shot gather in complex structural anisotropic model verify that the method is reliable and can be applied to the modelling of seismic wavefield in any anisotropic medium.
There are many ultra-low-speed layers, abnormal high pressure and vertical fracture in deep strata at Yinggehai Basin, which makes anisotropic and structure complex. The characteristics of seismic wavefield are complicated. Forward modelling is one of the main methods to study wavefield characteristics and observation system optimization. However, broad band and high frequency of the source wavelet in marine as well as the small grid size of the model result in a large amount of calculation. To better the calculation, the forward modelling of seismic wavefield in 3 D anisotropic media is developed based on GPU and CPU collaboration parallel in this paper. The 3 D model is divided and assigned to different processes, and the direction of parallel computing and the number of processes in each direction is arbitrarily selected, which can not only reduce the memory consumption in each process, but also reduce the calculation time. The forward modelling of seismic wavefield in a simple 3 D model verifies the method and can greatly improve the computational efficiency. The migration results of the synthetic common shot gather in complex structural anisotropic model verify that the method is reliable and can be applied to the modelling of seismic wavefield in any anisotropic medium.
引文
[1] 刘春成,陶杰,焦振华,等.海洋“犁式”电缆采集技术研究与实践[J].石油地球物理勘探,2016,51(6):1069-1074,1047.
[2] 刘春成,刘志斌,顾汉明.利用上/下缆合并算子确定海上上/下缆采集的最优沉放深度组合[J].石油物探,2013,52(6):623-629.
[3] 谢玉洪,李列,王大为,等.三维照明分析优化莺歌海盆地DF区海上三维地震采集观测系统[J].地质科技情报,2014,33(5):197-203.
[4] 司芗,吴国忱.裂隙等效TTI介质qP波反射特征研究[J].地球物理学进展,2012,27(5):2091-2099.
[5] Bohlen T.Parallel 3-D viscoelastic finite difference seismic modeling[J].Computers & Geosciences,2012,28(8):887-899.
[6] 周丽,顾汉明,成景旺,等.基于MPI的OBC三维多波多分量地震观测正演模拟并行算法实现[J].石油物探,2014,53(6):665-674.
[7] 桂生,刘洪,张玉洁.简化混合域全波形反演多GPU加速策略[J].地球物理学报,2017,60(2):665-677.
[8] 都志辉.高性能计算并行编程技术:MPI并行程序设计[M].北京:清华大学出版社,2001:5-7.
[9] 卢风顺,宋君强,银福康,等.CPU/GPU协同并行计算研究综述[J].计算机科学,2011,38(3):5-9.
[10] 蔡志成,顾汉明,成景旺,等.大规模变网格三维地震正演MPI并行策略与实现[J] .石油地球物理勘探,2017,52(3):468-476.
[11] 陈召曦,孟小红,郭良辉,等.基于GPU并行的重力、重力梯度三维正演快速计算及反演策略[J].地球物理学报,2012,55(12):4069-4077.
[12] 祝树云,朱旭光,颉冬莲,等.应用并行计算框架提升地震数据处理效率分析[J].石油地球物理勘探,2011,46(3):493-499.
[13] 李焱,胡祥云,杨文采,等.大地电磁三维交错网格有限差分数值模拟的并行计算研究[J].地球物理学报,2012,55(12):4036-4043.
[14] 宋国杰,杨顶辉,童平,等.求解3D弹性波方程的并行WNAD方法及其TI介质中的波场模拟[J].地球物理学报,2012,55(2):547-559.
[15] 张昆,董浩,严加永,等.一种并行的大地电磁场非线性共轭梯度三维反演方法[J].地球物理学报,2013,56(11):3922-3931.
[16] Nakata N,Tsuji T,Matsuoka T.Acceleration of computation speed for elastic wave simulation using a Graphic Processing Unit[J].Exploration Geophysics,2011,42(1):98-104.
[17] Weiss R M,Shragge J.Solving 3D anisotropic elastic wave equations on parallel GPU devices[J].Geophysics,2013,78(2):F7-F15.
[18] Zhang Y,Gao J.A 3D staggered-grid finite difference scheme for poroelastic wave equation[J].Journal of Applied Geophysics,2014,109(3):281-291.
[19] Yang P,Gao J,Wang B.A graphics processing unit implementation of time-domain full-waveform inversion[J].Geophysics,2015,80(3):31-39.
[20] Venstad J M.Industry-scale finite-difference elastic wave modeling on graphics processing units using the out-of-core technique[J].Geophysics,2016,81(2):29-37.
[21] 牟永光,裴正林.三维复杂介质地震数值模拟[M].北京:石油工业出版社,2005:81-82.
[22] Collino F,Tsogka C.Application of the perfectly matched absorbing layer model to the linear elastodynamic problem in anisotropic heterogeneous media[J].Geophysics,2001,66(1):294-307.
[2] 刘春成,刘志斌,顾汉明.利用上/下缆合并算子确定海上上/下缆采集的最优沉放深度组合[J].石油物探,2013,52(6):623-629.
[3] 谢玉洪,李列,王大为,等.三维照明分析优化莺歌海盆地DF区海上三维地震采集观测系统[J].地质科技情报,2014,33(5):197-203.
[4] 司芗,吴国忱.裂隙等效TTI介质qP波反射特征研究[J].地球物理学进展,2012,27(5):2091-2099.
[5] Bohlen T.Parallel 3-D viscoelastic finite difference seismic modeling[J].Computers & Geosciences,2012,28(8):887-899.
[6] 周丽,顾汉明,成景旺,等.基于MPI的OBC三维多波多分量地震观测正演模拟并行算法实现[J].石油物探,2014,53(6):665-674.
[7] 桂生,刘洪,张玉洁.简化混合域全波形反演多GPU加速策略[J].地球物理学报,2017,60(2):665-677.
[8] 都志辉.高性能计算并行编程技术:MPI并行程序设计[M].北京:清华大学出版社,2001:5-7.
[9] 卢风顺,宋君强,银福康,等.CPU/GPU协同并行计算研究综述[J].计算机科学,2011,38(3):5-9.
[10] 蔡志成,顾汉明,成景旺,等.大规模变网格三维地震正演MPI并行策略与实现[J] .石油地球物理勘探,2017,52(3):468-476.
[11] 陈召曦,孟小红,郭良辉,等.基于GPU并行的重力、重力梯度三维正演快速计算及反演策略[J].地球物理学报,2012,55(12):4069-4077.
[12] 祝树云,朱旭光,颉冬莲,等.应用并行计算框架提升地震数据处理效率分析[J].石油地球物理勘探,2011,46(3):493-499.
[13] 李焱,胡祥云,杨文采,等.大地电磁三维交错网格有限差分数值模拟的并行计算研究[J].地球物理学报,2012,55(12):4036-4043.
[14] 宋国杰,杨顶辉,童平,等.求解3D弹性波方程的并行WNAD方法及其TI介质中的波场模拟[J].地球物理学报,2012,55(2):547-559.
[15] 张昆,董浩,严加永,等.一种并行的大地电磁场非线性共轭梯度三维反演方法[J].地球物理学报,2013,56(11):3922-3931.
[16] Nakata N,Tsuji T,Matsuoka T.Acceleration of computation speed for elastic wave simulation using a Graphic Processing Unit[J].Exploration Geophysics,2011,42(1):98-104.
[17] Weiss R M,Shragge J.Solving 3D anisotropic elastic wave equations on parallel GPU devices[J].Geophysics,2013,78(2):F7-F15.
[18] Zhang Y,Gao J.A 3D staggered-grid finite difference scheme for poroelastic wave equation[J].Journal of Applied Geophysics,2014,109(3):281-291.
[19] Yang P,Gao J,Wang B.A graphics processing unit implementation of time-domain full-waveform inversion[J].Geophysics,2015,80(3):31-39.
[20] Venstad J M.Industry-scale finite-difference elastic wave modeling on graphics processing units using the out-of-core technique[J].Geophysics,2016,81(2):29-37.
[21] 牟永光,裴正林.三维复杂介质地震数值模拟[M].北京:石油工业出版社,2005:81-82.
[22] Collino F,Tsogka C.Application of the perfectly matched absorbing layer model to the linear elastodynamic problem in anisotropic heterogeneous media[J].Geophysics,2001,66(1):294-307.