用遗传算法进行瑞利波反演

瑞利波反演属于非线性最优化问题,已有的局部线性化方法(如阻尼最小二乘法)常使解估计陷入目标函数的局部极小值之中,且严重地依赖初始模型.另外,目前被广泛应用的一些简单的瑞利波反演方法也存在明显的缺陷.全局优化法-遗传算法大大放松了对初始模型选择的要求,且不易陷入局部最优解中.本文用已有的简单宜行的反演解释方法获取初始模型,从而确定模型参数的搜索范围,再用遗传算法反演得到最终的介质模型,效果非常理想.

THE APPLICATION OF GENETIC ALGORITHM TO RAYLEIGH WAVE INVERSION

The Rayleigh wave inversion is of the nonlinear optimization problem. The available local linearization techniques (such as the method of the damping least squares) oftencause the solution estimation to fall into the local minimum value of the objective function,and depend largely on the initial model. In addition, some simple Rayleigh wave inversionmethods widely used at present also have apparent defects.The overall optimization method-genetic algorithm-greatly lowers the requirements forselection of the initial modols and is unlikely to fall into the local optimal solution. Thepresent paper first used an available simple and convenient inverse interpretation method toobtain the initial model, thus deciding the searching range of model parameters, and thenmado inversion by using genetic algorithm to obtain the final medium model. The result isvery satisfactory.