地震资料波阻抗多尺度反演

Multiscale acoustic impedance inversion using poststack seismic data

对目标函数作修饰使它适应于用常规线性化方法求解，这一方面可以大大加快反演速度，另一方面又可以克服局部极值点的影响．使反演能够达到目标函数的全局最小点。近年来提出的多尺度反演方法．通过对目标函数作多尺度分解，化原来的非线性问题为一系列嵌套的线性问题，为非线性反演开辟了一条新途径。本文把这一思想用于地震资料波阻抗反演，由理论模型试算结果发现它确实能够改变反演性状，使求解结果跳出局部极值点，在理想情况下其收敛速度可比常规线性化反演快2倍。试验结果表明，起始分解尺度是影响反演效果的一个重要参数，其值应当与地震资料的主频相匹配。将该方法运用于实际资料处理取得了满意的结果。

Modifying the objective function makes it suitable for inversion with conventionallinearized algorithm. It is beneficial to both enhancing the inversion efficiency and preventing thefinal result from being trapped in local minimum. In recent years, multiscale inversion strategy inwhich original inversion problem is transformed into a series of cascade linearized problems hasshown as a promising approach to nonlinear inversion problems. In this paper the idea is applied toacoustic impedance inversion using poststack seismic data. The objective function is decomposed tomultiple scales with wavelet transform and hence the original inversion problem is re-formulated tobe a set of linearized problems corresponding to different scales. On each scale, the linearizedproblem is solved with the singular value decomposition method. The synthetic model test showsthat it has indeed changed the convergence character of the inversion and reduced the computingtime by nearly 2 /3 in comparison to that of conventional linear method. An exciting result has alsobeen obtained by applying it to field data.