CSAMT自适应正则化一维全资料反演

Adaptive regularized inversion of 1-D full CSAMT data

相比于传统反演算法对初始模型依赖性强的弱点,自适应正则化对初始模型要求较低,正则化因子可以根据每次迭代的数据目标函数和模型目标函数的关系,快速计算得到。这里采用自适应正则化反演算法尝试对一维全资料CSAMT的层状介质进行反演,通过三层和五层模型及其使用近场视电阻率数据的反演计算,表明了CSAMT自适应正则化一维全资料反演算法稳定,也发现该方法不仅能够避免传统CSAMT近场校正带来的误差,而且由于不用反复搜索正则化因子,加快了迭代收敛速度。

Compared with traditional inversion method, adaptive regularized inversion method does not depend on origi- nal model too much. Meanwhile regularized factor is calculated by the relation of the data object function and model object function which avoids constantly searching regularized factor and increasing speed. This paper adopts adaptive regularized in- version to discuss this method used in inversion of CSAMT layered strata. By the tests of 3 layers and 5 layers inversion vali- date this method is stable and useful for CSAMT.