非线性地球物理反演方法:回顾与展望

自20世纪90年代以来，非线性地球物理反演方法已走向成熟，这些方法包括线性化迭代法，仿真淬火法，遗传算法及联合反演方法等，线性化迭代反演基于开放物理系统状态发生相变的原理，要进一步改善模型参数化的技术和迭代过程的自组织；仿真淬火法与遗传算法基于自然过程的指数率或生物演化的优生率，可以相互结合以提高解估计的分辨率与置信度；联合反演要结合岩石物理性质的统计规律，才能取得兼容地质与综合方法的应用效果，地球动力学中的反问题不仅涉及偏微分方程系数项的求解，而且涉及初始条件或初始边界的求解，对地球动力学作用过程研究特别重要。

Non-linear Geophysical Inversion Methods: Review and Perspective

The Non linear geophysical inversion methods have became mature since 1990s. These methods include the linearized iteration method, the simulated annealing, the genetic algorithms and joint inversion methods. The linearized iteration method is based on the state changes of phases in an open physical system. It can be improved by advanced parameterization techniques and adoption of the self organization for the iterative processes. The simulated annealing and the genetic algorithms are based on exponentiation law of natural processes or optimal hybridizaton genetics. Combination of the advantages existent in both the simulated annealing and the genetic algorithms may generate even better inversion results by means of resolution and robustness. One should consider statistic equations linking different physical properties of rocks for the joint inversion in order to obtain better inversion results. Inversion problems of geodynamic processes involve not only with solving the coefficient terms in the partial differential equations, but also with solving the initial or boundary conditions, being a new challenge to the traditional geophysical inversion theories.