基于改进遗传算法的堆石体流变模型参数反馈分析

本文提出一种序列二次规划优化算法与标准遗传算法结合的流变模型参数反馈分析方法，这种算法既发挥了序列二次规划优化算法省时、高效、局部搜索能力强的特点，又发挥了遗传算法可以搜索到全局最优解而避免陷入局部极小值的优点，改善了常规遗传算法的收敛速度。将遗传算法搜索到的全局最优近似解作为初始值，代入收敛效率较高的序列二次规划程序进行最终局部优化。以某堆石坝为例，应用上述反演方法对高围压下的堆石体9参数流变模型参数进行了反演分析，验证了此方法的可行性与有效性。

Back analysis of rockfill creep model parameters based on improved genetic algorithm

An intelligent back analysis method based on genetic algorithm(GA) and sequence quadratic programming(SQP) is proposed in this paper.GA with group search and global convergence can efficiently overcome the problem of high sensitivity to initial guess and not run into local minimum.SQP has a high convergence rate and precision solution for local search.A hybrid genetic algorithm(HGA) which combines the advantages of GA and classical SQP algorithm is presented and overcomes the shortcoming of GA,namely showing the lower convergence rate at the time near true solution.The global convergent approximate solution searched by GA is accepted as the initial input value with SQP optimizing analysis to gain the final optimal solution.HGA is applied to perform back analysis of a rockfill creep model with nine parameters under high confining pressure.The results show that the computed values of rockfill creep deformation fit closely to actual monitoring data.