中文摘要：

为了得到准确的土壤结构,提出了一种基于遗传算法（genetic algorithm,GA）的土壤结构逆问题求解方法,该方法通过视在电阻率测量值与GA产生的土壤结构对应视在电阻率计算值的差异,从而得到一个最优的土壤结构。与现有的利用GA求解土壤结构逆问题的方法相比,该方法的优点在于首先合理的设置了土壤参数初始精度,压缩了求解空间,使得简单土壤结构的求解能直接收敛到该精度下的最优;同时,对于复杂土壤结构求解,还引入了遗传–灾变算法与变精度的方法。因而在求解土壤结构时,解空间会被限制在一个合适的规模,同时也会引入最优解附近新的个体,使迭代求解过程更容易跳出局部解,从而逼近全局最优。计算结果表明,两层土壤结构可以收敛到全局最优,适应度函数值Fg=0,三层土壤结构下,该方法求解的Fg要小于标准遗传算法;同时,该方法输出稳定,10次重复计算下的输出结果差异小。

英文摘要：

In order to get an accurate soil structure, we put forward a method for solving the soil structure inverse problem by genetic algorithm（GA）, in which the distinction produced by the experimental and theoretical curves of soil apparent resistivity is adopted to achieve an optimized soil structure. Compared with the present method which is used to solve the soil structure based on GA, first, the proposed method possesses advantages that we rationally set the initial precision of soil parameters, then the solution space is compressed, which makes the inverse problem of simple soil structure to directly converge to optimum in this precision; meanwhile, for the inverse problem of complex soil structure, the genetic-reckoning algorithm and variable precision method are introduced. So, during the estimation of soil structure, the solution space will be compressed into a certain scale and new individuals around the optimal solution will be introduced, which make the iteration process to more easily escape the local optimal solution, approaching the overall optimum. The calculation shows that the results of two-layer soil structure can coverage to overall optimum, the fitness value Fg=0 and for three-layer soil structure, the Fg calculated by the paper is smaller than simple GA; at the same time, the proposed method has a more stable output, the output difference is small in ten times repeated calculation.