中文摘要：

Monte Carlo（MC）法在目前边坡可靠度分析中是一种相对精确的方法，应用广泛，受问题限制的影响较小，适应性很强，其误差仅与标准差和样本容量有关。但其精度受随机抽样的可靠性和模拟次数制约，收敛速度慢，影响了实际使用。在极限平衡方法的基础上，用拉丁超立方抽样（Latin hypercube sampling，LHS）方法代替MC法的随机抽样，考虑边坡参数的变异性和相关性进行边坡可靠度分析。讨论了LHS法、MC法中可靠指标的各种计算方法，建议以破坏概率、安全系数均值和标准差作为评价指标。算例显示LHS法较MC法效率上有很大改善：较少的抽样样本就能反映参数的概率分布，可靠度分析收敛快，不需要大量的模拟，因此，值得在边坡可靠度分析中推广应用。也将工程上常用的均匀设计和正交设计用于边坡可靠度分析，结果表明，正交设计结果和中心点法比较接近，而均匀设计得到的结果则是不可靠的。

英文摘要：

Monte Carlo （MC） simulation has been widely applied to the slope reliability analysis and is accepted as a precise method, which is applicable for complex problems and its error is only affected by standard deviation and sample size. But its precision is affected so much by the reliability of sampling and the simulation times. Based on the limit equilibrium method, the Latin hypercube sampling（LHS） technique is employed instead of the MC method to analyze the slope reliability accounting for the parameters variances and correlations. Then the slope reliability index and failure probability can be computed. The methods for computation of the reliability index in LHS or MC method are discussed, and failure probability together with mean and standard deviation of factor of safety are recommended as evaluation index. The results of example problems show that LHS method is more efficient than MC simulation, which can strongly improves the representation of soil parameters and tends to convergence within relative few simulation times. Hence, the LHS method should be extended in practical slope reliability analysis. The uniform design and orthogonal design are applied to slope stability analysis. The results of orthogonal design is close to FOSM but results of uniform design are unreliable.