单斜介质中方位NMO速度Thomsen参数反演方法研究

采用启发式共轭梯度法，即随机爬山法 + 共轭梯度法，利用单斜介质中P波方位NMO速度椭圆轴向偏转角度接近于零这一特性，简化P波方位NMO速度公式，并利用多方位P波NMO速度，反演出某一初始CMP观测线与自然坐标系之间的夹角，作为进一步进行Thomsen各向异性参数反演的基础. 根据各向异性介质中方位NMO速度与Thomsen参数之间的关系，建立了利用三种波的多方位NMO速度及垂直传播速度反演单层单斜各向异性介质Thomsen各向异性参数的目标函数. 对计算的理论值添加具有一定标准差的正态分布的随机噪声，用以模拟实际观测存在的误差，通过对加噪后的数据进行多次反演的误差分析，表明了所建立的目标函数及选用的反演方法是有效可行的，而且相对稳定.

Thomsen parameter inversion with azimuthal NMO velocity in monoclinic media

With the combination of the heuristic and in_heuristic methods, we have studied a hybrid algorithm, that is stochastic hill_climbing conjugate gradient. We simplified the P_wave NMO ellipse equations according to the fact that the rotation angle of the NMO ellipses for P_wave is close to zero in monoclinic media. So, at first, we performed inversion of angle between the initial CMP line and the natural coordinate as the base of the subsequent inversion of Thomsen's anisotropic parameters. Then the objective functions are established for joint inversion of Thomsen's anisotropic parameters with vertical velocities and NMO ellipses of three pure modes in a single monoclinic layer. To simulate errors in measured data, Gaussian noise with a variance to certain extent is added to vertical and NMO velocities. Then the inversion procedure is carried out for each realization of the data distorted by noise. The result of error analysis confirmed the accuracy and stability of the objective function and inversion method.