论文部分内容阅读
针对传统差分演化算法在反演瑞雷波频散曲线中存在交叉概率和缩放因子选取不当导致反演结果不准确等问题,在传统差分演化算法的基础上,把控制参数直接编码到个体中,即采用参数自适应差分演化算法对高频瑞雷波频散曲线进行反演来获得近地表的横波速度结构。合成和实际数据的反演实验结果表明:参数自适应差分演化算法继承了传统差分演化算法简单、高效的特点同时,还可在频散曲线的反演中自动地选取合适的参数值并正确地进行反演迭代,无需再通过试验获得交叉概率和缩放因子两个控制参数;频散曲线反演中目标函数的收敛性好,改进算法在迭代的过程中能够快速收敛到全局最优;模型参数的概率分布高,即在大的空间搜索范围内,参数自适应差分演化算法仍然能够准确地搜索到真值范围并找到全局极小值,保证了反演的结果可靠度,使其能有效地应用于瑞雷波频散曲线的反演和解释。
Aiming at the problems of traditional differential evolution algorithm such as crossover probability and inaccuracy of inversion factor in the inversion Rayleigh wave dispersion curve, the control parameters are directly encoded into individuals based on the traditional differential evolution algorithm, That is to say, the near-surface S-wave velocity structure is obtained by inverting the dispersion curve of the high-frequency Rayleigh wave using the adaptive differential evolution algorithm. The experimental results show that the adaptive differential evolution algorithm inherits the simple and efficient characteristics of the traditional differential evolution algorithm and can automatically select the appropriate parameter values in the inversion of the dispersion curve and correctly For the inversion iteration, the two control parameters of crossover probability and scaling factor need not be obtained through experiments. The convergence of the objective function in the dispersion curve inversion is good. The improved algorithm can converge to the global optimum rapidly in iterative process. The model parameters The probability distribution is high, that is, in the large space search range, the parameter adaptive differential evolution algorithm can still search the true value range accurately and find the global minimum value, which ensures the reliability of the result of inversion so that it can effectively Applied to Rayleigh wave dispersion curve inversion and interpretation.