复杂近地表三维初至波走时层析方法研究

3D first-arrival travel-time tomography based on complex near-surface model

复杂起伏地表条件下三维初至波走时层析速度建模方法是我国西部山地、沙漠地区地震资料处理的关键技术之一.传统的三维走时层析反演在应用中存在诸多问题:一是射线追踪技术固有的计算效率低、对复杂模型计算不稳定;二是对于大规模三维模型,Tikhonov正则化难以对零空间和欠定分量进行有效约束,造成迭代收敛速度缓慢.本文首先在多模板快速推进算法(MSFM)走时计算的基础上,提出了一种新的射线追踪方法.整形正则化方法和共轭梯度法对反演方程进行了有效的约束,实现了初至波走时层析反演.三维理论模型实验和实际资料处理表明,该方法具有比传统射线走时层析方法更高的反演精度与迭代收敛速度.

Highly efficient and accurate ray-tracing method in complex near-surface media have continuously been the key technique in three dimensional travel-time tomography. However, traditional algorithms utilized for travel-time tomography usually encounter problems when handling large 3 D seismic data. First is the shortcomings inherited in the ray-tracing technique,such as the low computational efficiency and instability when strong velocity contrast exists in velocity model. Second,the disadvantages of traditional regularization algorithms lie in their low iterative convergence rate caused by the insufficient control on estimated model by the Tikhonov＇s regularization,which could be detrimental for large 3 D scale problems,when only few iterations are affordable. In this paper,we first develop the Multi-Stencils Fast Marching Method（ MSFM） into 3 D travel-time computation for complex media with topography. We propose a new algorithm for calculating the Fréchet derivative matrix based on MSFM. Additionally,we incorporate a shaping regularization into conjugate gradient algorithm to minimize the data misfit iteratively. Finally,the numerical experiments on 3 D models and real data have demonstrated the higher accuracy and faster convergence of the proposed method compared to the classical algorithm.