摘要
提高土质边坡稳定可靠度计算的精度和效率对工程设计具有重要的意义,为此提出了基于改进Kriging代理模型的土质边坡稳定可靠度分析方法。首先采用基于模拟退火的粒子群算法优化Kriging模型相关参数,并以一典型算例验证了该模型的良好函数拟合能力。接着将改进的Kriging模型作为边坡极限状态函数的代理,通过拉丁超立方抽样获得样本点,并由有限差分强度折减法计算样本点对应的响应值,利用基于熵理论的学习函数逐步更新代理模型,直至达到精度要求,进而基于该模型通过蒙特卡洛模拟评估边坡失效概率。最后,以一双层土质边坡问题为例,研究了所提方法的有效性。结果表明:所提方法不仅能够逐步提高计算精度,且极大地降低了计算量,为复杂边坡稳定可靠度问题提供了一种高效的评估方法。
Improving the accuracy and efficiency of reliability calculation of soil slope stability is of great significance for engineering design.Therefore,a reliability analysis method of soil slope stability based on improved Kriging surrogate model is proposed.Firstly,the related parameters of Kriging model are optimized by particle swarm optimization based on simulated annealing,and a typical example is given to verify the good function fitting ability of the model.Then,the improved Kriging model is used as the surrogate of the limit state function of a slope.Sample points are obtained by Latin hypercube sampling,and the corresponding response values of sample points are calculated with finite difference strength reduction method.The surrogate model is updated step by step by using learning function based on entropy theory.When this model reaches the accuracy requirement,it stops updating and then evaluates the failure probability of slope by Monte Carlo simulation based on the model.Finally,the effectiveness of the proposed method is studied by taking a two-layer soil slope problem as an example.The results show that the proposed method can not only gradually improve the accuracy of calculation,but also greatly reduce the amount of calculation.It provides an efficient evaluation method for complex slope stability reliability problems.
作者
邓志平
潘敏
牛景太
DENG Zhiping;PAN Min;NIU Jingtai(National and Provincial Joint Engineering Laboratory for the Hydraulic Engineering Safety and Efficient Utilization of Water Resources of Poyang Lake Basin,Nanchang Institute of Technology,Nanchang 330099,China)
出处
《南昌工程学院学报》
CAS
2019年第6期92-97,共6页
Journal of Nanchang Institute of Technology
基金
江西省教育厅科学技术研究项目(GJJ180958,GJJ161096)
国家自然科学基金资助项目(51769017)
