基于改进遗传算法的边坡可靠度分析

Slope reliability analysis based on improved genetic algorithm

采用基于概率的可靠性分析方法来分析边坡的稳定性,以克服目前大多数分析方法以确定性安全系数为评判标准的弊端。对于每次搜索中给定的滑移面,结合拉格朗日乘子法提出边坡可靠度的计算模型;采用遗传算法搜索关键滑移面,提出边坡的适应度函数。为克服标准遗传算法在进化过程中无法调节全局搜索和局部搜索速度、且可能陷入次优解的缺陷,改进了子代中最佳个体的确定方法,来并以相邻两代的群体多样性特征值增量为导向,动态调节交叉概率和变异概率,以保持群体的多样性和搜索的有效性。实例分析结果表明,本文方法可以提高遗传算法的搜索能力和收敛速度,保证解全局最优;该法适用于边坡稳定可靠度分析,且前处理简单;在遗传操作前预优化可靠指标计算模型能大大减少遗传算法中的计算量,节约机时。

Reliability analysis methods based on probability is proposed for analyzing slope stability to overcome the shortcomings caused by using safety factor as the standard of slope appraisal. For each given slip surface, an optimization model is suggested combined with Lagrange multiplier method. Genetic algorithm （GA） is adopted to search for the key slip surface, and a fitness function is proposed. According to the disadvantages of standard GA, the global and local research abilities can＇t be controlled in the optimization process; the method of determining the best individual in offspring is improved, and the probabilities of crossover and mutation are adjusted dynamically guided by the increment of diverse characteristic value to assure population diversity and search validity. The results show that the search ability and convergence speed are improved, and the global optimal solution is guaranteed by the proposed method; the method is suitable for analyzing the reliability of slope stability with simple pretreatment; computational complexity of GA can be reduced greatly and much calculation time can be saved if computational model of reliability index is pre-optimized before GA performance.