土石坝边坡小概率失效计算方法

针对蒙特卡罗随机有限元方法求解土石坝边坡小概率失效问题计算时间过长这一弊端,提出了基于链表筛分法和Kriging代理模型的混合子集模拟法。首先,利用链表筛分法计算初始样本点,并以初始样本点作为Kriging代理模型的训练样本点。然后在子集模拟法的分层模拟中采用Kriging代理模型预测条件样本点的响应值,从而可计算出土石坝边坡失效概率。最后,通过一个土石坝边坡算例来说明本方法相比其他方法所需要的计算时间更少,更适合计算土石坝边坡小概率失效问题。

Calculation of small failure probability in slopes of earth dams

A hybrid subset simulation approach based on the linked-list sifting method and Kriging surrogate model is proposed to overcome the drawback that it takes long time for the Monte-Carlo-based stochastic finite element method to calculate small failure probability in slopes of earth dams. First, the linked-list sifting method is adopted to calculate initial sample points that can be used for training the Kriging surrogate model. Next,the Kriging surrogate model is used to predict the response values of conditional sample points at different levels of subset simulation, which are used to determine the slope failure probability. Finally, a slope example of an earth dam is given to illustrate that the proposed approach costs less time than other approaches, which is more suitable to calculate small failure probability in slopes of earth dams.