一种多信息约束的初至波走时层析反演优化方法

初至波走时层析成像是准确获取复杂近地表表层速度结构的有效方法。对西部复杂山前带而言, 近地表速度建模精度低, 严重影响后续中深层速度建模和偏移成像质量。因此, 研究一种高效、可适应复杂近地表条件的初至波走时层析算法非常重要。提出了一种基于Tikhonov正则化方法多信息约束的初至波走时层析反演的优化策略, 通过多模板快速推进算法(MSFM)解程函方程计算网格单元走时, 用迭代反演方法线性化计算非线性逆问题, 避免大规模Frechet矩阵的求取和存储, 提升了计算效率; 根据微测井信息建立初始速度模型, 利用视慢度等先验信息对层析反演目标函数进行约束, 采用Tikhonov正则化来解决最小化数据差和模型平滑度的反问题, 降低了目标方程的病态性, 从而提高反演精度。起伏地表模型和西部实际资料应用结果表明, 该方法通过对初至走时和走时-偏移距曲线联合拟合有效适应复杂地表条件, 解决了三维地震资料初至波走时层析成像出现的“高速异常体”问题, 提高了反演的稳定性和计算精度。

First-arrival travel-time tomography is an effective method for accurately calculating complex near-surface velocity structures.However, it is difficult to apply the traditional first-arrival travel-time tomography algorithm to seismic data prospecting with large data volumes, owing to the need for excessive memory and calculation time.The accuracy of near-surface velocity modeling of super-large offset data and complex piedmont data in western China is insufficient, which seriously affects the quality of subsequent mid-deep modeling and migration imaging.Therefore, it is very important to study first-arrival travel-time tomography for complex near-surface schemes with high computational efficiency and accuracy.In this study, we present an optimal first-arrival travel-time tomography method with multi-information constraints.The underlying idea is to introduce prior information into the objective function through the regularization method and improve the calculation efficiency through the optimization of forward and inversion algorithms.We use the multi-stencil fast marching method (MSFM) to solve the Eikonal equation to calculate the travel-time of grid elements and use an iterative inversion method to linearize the nonlinear inverse problem.The optimal algorithm can improve the accuracy of the first-arrival travel-time calculation of complex near-surface structures and avoid large-scale Frechet matrix calculation and storage, which greatly improves the computational efficiency.In terms of objective function constraints, we adopt Tikhonov regularization to solve the inverse problem of minimizing data difference and model smoothness and increase prior information such as apparent slowness to constrain the solution space of the model.The traditional objective function only fits the travel time of the first-arrival wave; therefore, it is difficult to distinguish the receiving points with the same travel-time and different positions.When dealing with complex near-surface problems, it is easy to observe the phenomenon of "bull's eye (high-speed abnormal body)".The objective function combined with apparent slowness information can be used to fit the first-arrival travel-time and travel-time offset curves, which can effectively distinguish the receiving points with the same travel-time and different positions and achieve better results for near-surface problems.Adding slowness to the regularization operator to normalize the model can also avoid the problem of decreasing the resolution of inversion in the high-speed region.These methods reduce the ill-conditioning of the target equation and improve the inversion accuracy.In addition, the initial velocity model based on micrologging information can make the inversion result closer to the real velocity structure.The undulating surface model and actual data processing show that this method can effectively solve the problem of "high-speed abnormal bodies in the first-arrival travel-time tomography of 3D seismic data.Compared with traditional first-arrival travel-time tomography, our surface travel-time tomography improves inversion stability and calculation accuracy.