基于拉丁方抽样及K-S检验的边坡可靠性分析

Reliability analysis of slope with Latin hypercube sampling and K-S test

边坡稳定性是受复杂因素影响下的多维非线性问题,大多数边坡工程都存在着不确定性。大量试验和工程实践证明,影响边坡状态的因素中有许多具有显著的随机性,参数具有变异性。从概率的角度出发,结合实际工程,对于原始数据采用K-S检验法进行假设验证,确定参数变量的分布类型,克服了人为假设的误差,并分别采用Latin方抽样(LHS)法和Monte Carlo法对参数进行抽样,得到状态函数值,确定安全系数及可靠指标,对比两种抽样方法,基于LHS法模拟次数要少于Monte Carlo法,而且破坏概率的收敛性也优于Monte Carlo法,明显节省了计算时间。

Slope stability is the multi-dimensional nonlinear problem which is influenced by complex factors. There is uncertainty in most of slope engineering. A large number of test and engineering practice have proved that the factors affecting slope state have significant randomness and parameters have variability. So we can use Kolmogorov-Smirnov test （K-S test） for the raw data to determine the distribution type of parametric variation from the perspective of probability and practical engineering. It will overcome artificial assumption. And sample the parameters to obtain state function and determine the safety factor and reliability index using Latin hypercube sampling （LHS） method and Monte Carlo method, respectively. Comparison of two sampling methods, the number of simulation based on LHS is less than the Monte Carlo method. Also the convergence of failure probability curve with LHS is better than the Monte Carlo method.