中文摘要：

为了构建合理可靠的初始模型，使反演稳定快速收敛且达到更高的精度.本文提出用控制转换算法构建初始模型参数，然后使用最优分割法优化初始模型.一维MT最小二乘反演计算结果表明：该方法构建的初始模型使反演稳定快速收敛；使用相同的反演方法，该方法构建的初始模型的反演迭代次数比Bostick法构建的初始模型的反演迭代次数少20%以上，比均匀半空间作为初始模型的反演迭代次数少100%以上，并且反演精度也有了提高.该方法不但对提高MT一维反演效果有意义，而且对构建其他勘探方法的反演初始模型也具有一定的参考价值.

英文摘要：

In order to build reasonable and reliable initial model, make inversion is steady and fast convergence, this paper proposes a new method of building initial model that builds initial model parameters by control transformation algorithm and optimizes initial model by optimal segmentation. The calculation results of 1D MT least-squares inversion show that this method makes inversion is steady and fast convergence. For the same inversion method, the number of inversion iteration based on this method of building initial model reduces by more than 20% than the result of initial model built by Bostick inversion and reduces by more than 100% than the result of the initial model of homogeneous half space, and the accuracy of inversion is improved. This method not only is meaningful for improving the effect of MT 1D inversion, but also has reference value for the inversion of other exploration methods.