Comparison of four optimization methods in elastic full-waveform inversion
LIU Yuhang1, HUANG Jianping1,2, YANG Jidong1,2, LI Zhenchun1,2, KONG Linghang1, DING Zhaoyuan3
1. School of Geosciences, China University of Petroleum(East China), Qingdao, Shandong 266580, China; 2. Pilot National Laboratory for Marine Science and Technology(Qingdao), Qingdao, Shandong 266071, China; 3. Tarim Oilfield Company, PetroChina, Korla, Xinjiang 841000, China
Abstract:Elastic full-waveform inversion (EFWI) is a high-precision imaging method. Since it is a highly nonlinear problem in nature, local optimization algorithms are often used to solve it, but the inversion results of different optimization algorithms are very different. On the basis of the commonly used conjugate gradient (CG) method and limited-memory Broyden-Fletcher-Goldfarb-Shanno (L-BFGS) method, the pseudo-Hessian matrix is adopted as the gradient preconditioning operator to devise a preconditioning CG (P-CG) method and a preconditioning L-BFGS (P-L-BFGS) method. The principles and implementation processes of these four optimization algorithms are expounded. Then, the diffraction model and the Marmousi Ⅱ model are used to test the four algorithms. The following conclusions can be drawn from the results:①The pseudo-Hessian preconditioning operator can compensate for the deep energy and accelerate the convergence of inversion; ②The CG method and the P-CG method are easy to implement, but they cannot suppress the multi-parameter coupling effect because they only use first-order gradients. However, the P-CG method can deliver inversion results that are slightly inferior to those of the L-BFGS method for the complex Marmousi Ⅱ mo-del. ③The implementation of the L-BFGS method and the P-L-BFGS method is more complicated. Nevertheless, because the pseudo-Hessian matrix is calculated in the inversion process, the multi-parameter coupling effect is suppressed to a certain extent. ④As for the Marmousi Ⅱ model, both the L-BFGS method and the P-L-BFGS method can obtain compressional wave (P-wave) and shear wave (S-wave) velocity models with high inversion accuracy, but density inversion will be subject to the overfitting phenomenon.
JEONG W,LEE H Y,MIN D J. Full waveform inversion strategy for density in the frequency domain[J]. Geophysical Journal International,2012,188(3):1221-1242.
[2]
刘张聚,童思友,方云峰,等. 基于时域加权的拉普拉斯-频率域弹性波全波形反演[J]. 石油地球物理勘探,2021,56(2):302-312,331.LIU Zhangju,TONG Siyou,FANG Yunfeng,et al. Full elastic waveform inversion in Laplacian-Fourier domain based on time-domain weighting[J]. Oil Geophysical Prospecting,2021,56(2):302-312,331.
[3]
王华忠,盛燊.走向精确地震勘探的道路[J].石油物探,2021,60(5):693-708.WANG Huazhong,SHENG Shen. Pathway toward accurate seismic exploration[J]. Geophysical Prospecting for Petroleum,2021,60(5):693-708.
[4]
VIGH D,SUN D,WATTS D,et al. Elastic full-waveform inversion application using multicomponent measurements of seismic data collection[J]. Geophysics,2014,79(2):R63-R77.
[5]
邵祥奇,何兵寿,史才旺. 基于分频编码的弹性波全波形反演[J]. 石油地球物理勘探, 2020, 55(6):1321-1329.SHAO Xiangqi,HE Bingshou,SHI Caiwang. Elastic full waveform inversion based on frequency-division encoding[J]. Oil Geophysical Prospecting,2020,55(6):1321-1329.
[6]
王毓玮,董良国,黄超,等. 降低弹性波全波形反演强烈非线性的分步反演策略[J]. 石油地球物理勘探,2016,51(2):288-294.WANG Yuwei,DONG Liangguo,HUANG Chao,et al. Stepwise inversion strategy for reducing strong nonlinearity in full waveform inversion of elastic wave[J]. Oil Geophysical Prospecting,2016,51(2):288-294.
[7]
VIRIEUX J,OPERTO S. An overview of full-waveform inversion in exploration geophysics[J]. Geophysics,2009,74(6):WCC1-WCC26.
[8]
唐焕文. 实用最优化方法[M]. 辽宁大连:大连理工大学出版社,2004.
[9]
TARANTOLA A. Inversion of seismic reflection data in the acoustic approximation[J]. Geophysics,1984,49(8):1259-1266.
[10]
WANG Z Y,ZHANG Y N,CAI H M,et al. Nonli-near process control of wave-equation inversion and its application in the detection of gas[J]. Journal of Computer Applications,2008,28(3):3-13.
[11]
马昌凤. 最优化方法及其Matlab程序设计[M]. 北京:科学出版社,2010.
[12]
BROSSIER R. Two-dimensional frequency-domain vi-sco-elastic full waveform inversion:Parallel algorithms,optimization and performance[J]. Computers and Geosciences,2010,37(4):444-455.
[13]
MALINOWSKI M,OPERTO S,RIBODETTI A. High-resolution seismic attenuation imaging from wide-aperture onshore data by visco-acoustic frequency-domain full-waveform inversion[J].Geophysical Journal International,2011,186(3):1179-1204.
[14]
张凯,李振春,陈永芮,等. 一种新的用于全波形反演的能量加权梯度方法[J]. 石油物探,2015,54(1):77-82.ZHANG Kai,LI Zhenchun,CHEN Yongrui,et al. A new energy weighted gradient method for full waveform inversion[J]. Geophysical Prospecting for Petroleum,2015,54(1):77-82.
[15]
GAUTHIER O,VIRIEUX J,TARANTOLA A. Two-dimensional nonlinear inversion of seismic waveforms:Numerical results[J]. Geophysics,1986,51(7):1387-1403.
[16]
PRATT R G,SHIN C,HICK G J. Gauss-Newton and full Newton methods in frequency-space seismic waveform inversion[J]. Geophysical Journal International,1998,133(2):341-362.
[17]
SHIN C S,JANG S H,MIN D J. Improved amplitude preservation for prestack depth migration by inverse scattering theory[J]. Geophysical Prospecting,2001,49(5):592-606.
CHEN K,SACCHI M D. Elastic least-squares reverse time migration via linearized elastic full-waveform inversion with pseudo-Hessian preconditioning[J]. Geophysics,2017,82(5):S341-S358.
[20]
PLESSIX R E. A review of the adjoint-state method for computing the gradient of a functional with geophysical applications[J]. Geophysical Journal International,2006,167(2):495-503.
[21]
TARANTOLA A. A strategy for nonlinear elastic inversion of seismic reflection data[J]. Geophysics,1986,51(10):1893-1903.
[22]
FORGUES E,LAMBARE G. Parameterization study for acoustic and elastic ray+Born inversion[J]. Journal of Seismic Exploration,1997,6(2):253-278.
[23]
KÖHN D,DE NIL D,KURZMANN A,et al. On the influence of model parametrization in elastic full waveform tomography[J]. Geophysical Journal International,2012,191(1):325-345.
[24]
郭振波. 弹性介质波形反演方法研究[D]. 山东青岛:中国石油大学(华东),2014.
[25]
张广智,姜岚杰,孙昌路,等. 基于照明预处理的分步多参数时间域声波全波形反演方法研究[J]. 石油物探,2017,56(1):31-37.ZHANG Guangzhi,JIANG Lanjie,SUN Changlu,et al.The stepped multi-parameter FWI of acoustic media in time-domain by L-BFGS method with illumination analysis[J]. Geophysical Prospecting for Petroleum,2017,56(1):31-37.
[26]
HESTENES M R,STIEFEL E L. Methods of conjugate gradients for solving linear systems[J]. Journal of Research of the National Bureau of Standards(United States),1952,49(6):409-436.
[27]
FLETCHER R,REEVES C M. Function minimization by conjugate gradients[J]. The Computer Journal, 1964,7(2):149-154.
[28]
KÖHN D. Time Domain 2D Elastic Full Waveform Tomography[D]. Kiel University,Schleswig-Holstein,Germany,2011.
[29]
DAVIDON W C. Variable metric method for minimi-zation[J]. Siam Journal on Optimization,1991,1(1):1-17.
[30]
NOCEDA J,WRIGHT S J. Numerical Optimization(2nd Edition)[M]. Springer Science & Business Media,New York,USA,2006.
[31]
苗永康. 基于L-BFGS算法的时间域全波形反演[J].石油地球物理勘探,2015,50(3):469-474.MIAO Yongkang. Full waveform inversion in time domain based on L-BFGS algorithm[J]. Oil Geophy-sical Prospecting,2015,50(3):469-474.
[32]
FU L,GUO B W,SCHUSTER G T. Multiscale phase inversion of seismic data[J].Geophysics,2018,83(2):R159-R171.
[33]
NGUYEN B D,MCMECHAN G A. Five ways to avoid storing source wavefield snapshots in 2D elastic prestack reverse time migration[J]. Geophysics,2015,80(1):S1-S18.