黄鑫泉,
万先武,
孙敏傲,
董良国
同济大学海洋地质国家重点实验室, 上海 200092
基金项目: 国家自然科学基金（41774122，41630964，41674115）、国家重点研发计划（2018YFC0310100）与中国科学院战略性先导科技专项（A类）（XDA14010203）共同资助
详细信息
作者简介: 刘玉柱, 男, 博士, 1979年生, 教授、博士生导师, 主要从事地震波正反演理论、方法与应用研究.E-mail:liuyuzhu@tongji.edu.cn
中图分类号: P631;P315 收稿日期:2018-08-13
修回日期:2018-09-10
上线日期:2019-05-05
Elastic multi-parameter full-waveform inversion for anisotropic media
LIU YuZhu,HUANG XinQuan,
WAN XianWu,
SUN MinAo,
DONG LiangGuo
State Key Laboratory of Marine Geology, Tongji University, Shanghai 200092, China
MSC: P631;P315
--> Received Date: 13 August 2018
Revised Date: 10 September 2018
Available Online: 05 May 2019
摘要
摘要:各向异性介质弹性波方程全波形反演过程中多参数之间的相互耦合，使得弱参数在反演过程中难得到理想的结果.本文以VTI介质为例，在各参数辐射模式分析的基础上，基于改进的散射积分算法实现目标函数梯度的直接求取，进一步构建高斯牛顿方向，实现Hessian矩阵的有效利用，以考虑Hessian矩阵非主对角线元素包含的各参数间的耦合效应，在不使用任何反演策略的情况下实现高精度的VTI介质弹性波方程多参数同步反演.同时，该方法在计算过程中无需存储庞大的核函数矩阵，且无需传统截断牛顿法中额外的正演计算，因此内存占用小，计算效率高.本文数值试验验证了该方法的有效性，为各向异性多参数全波形反演提供了一种新的解决方案.
关键词: 全波形反演/
各向异性/
弹性波方程/
多参数/
高斯-牛顿方向/
改进的散射积分算法
Abstract:The trade-off effects between parameters in anisotropic media, which have been studied in radiation patterns analysis, makes it difficult to get acceptable weak parameter results during full-waveform inversion (FWI) based on an elastic wave equation. In this paper, to reduce the trade-off effects between parameters, we apply the truncated Gauss-Newton method to anisotropic FWI, taking VTI media for example. All parameters have been revealed simultaneously without using any strategy. Based on the improved scattering-integral (SI) algorithm, the steepest-descent direction can be directly calculated by vector operations, without storing the huge Fréchet kernels in memory beforehand. The Gauss-Newton descent direction iteration needs just two vector operations without any extra forward computations. Numerical experiments in the final part prove the effectiveness of the proposed method which opens a new approach on simultaneously constructing anisotropic parameters in FWI.
Key words:Full-waveform inversion/
Anisotropy/
Elastic wave equation/
Multi-parameter/
Gauss-Newton method/
Improved scattering-integral algorithm
PDF全文下载地址:
http://www.geophy.cn/data/article/export-pdf?id=dqwlxb_14979
