叠后地震道反演的随机爬山法

Stochastic hillclimbing algorithm for poststack seismic trace inversion

本文在Huang等人（1995）提出的随机爬山法（SHC）基础上，使用Ingber（1993）给出的依赖于温度的似Cauchy分布产生的新模型，构造一种叠后地震道及演的新随机爬山法。此法的基本思路是将目标函数定义为归一化的相似系数E，通过SHC法反演得到一个波阻抗模型使E达到极大值。此法在每次迭代构造新模型时，既能在高温时进行稀疏点搜索，又能在低温时于当前模型进行搜索，从而加快了迭代和收敛速度。文中通过对实际声波测井曲线进行SHC反演，采用有、无噪声两种情况，经过44次迭代，相似系数E达到0．99，不论有、无噪声，反演结果均与实际曲线非常接近。该法不仅可以反演出高分辨率的波阻抗剖面，而且可以用于反演孔隙率参数。

A new st0chastic hillclimbing (SHC) algorithm for poststack seismic trace in-version can be formed by using both the stochastic hillclimbing algorithm of Huanget al in l995 and the new model that was derived from Ingber's temperature depen-dent pseudo-Cauchy distribution (1993). The basic principle of the method is thatthe objective function is defined as normalized similarity coefficient E, which is thenmade maximum by the wave impedance model derived from SHC inversion. In eachnew model construction using iteration, both the sparse-point search at high tem-perature and the present model search at low temperature can be conducted, thusspeeding up iteration and convergence.In SHC inversion of real acoustic logging data with noises and without noise,the similarity coefficient E is up to O. 99 after 44 iterations; and whether there isnoise or no, the inversion results are very similar to real curves. The method can beused to produce high-resolution wave impedance section and porosity.