This study investigates the effect of spatial variability on the stability of a real 3D slope subjected to earthquakes by conventional Monte Carlo simulation (MCS) and a stochastic response surface method. The stochastic response surface method used in this study (VRSRSM) was combined with a variance reduction method due to spatial averaging of soil parameters in slope failure to reduce a large computational load due to the realization of spatially correlated input parameters in three-dimensional space. In order to verify the applicability of the stochastic response surface method with a variance reduction method for 3D slope reliability, the safety factor for a simple 3D slope under non-seismic and seismic conditions was analyzed by MCS and VRSRSM. The VRSRSM was applied to risk evaluation of deep-seated landslides caused by the 2016 Kumamoto earthquake along with backcalculation of strength parameters based on the slope failure geometry and field investigation.

本研究通过传统的蒙特卡罗模拟 (MCS) 和随机响应面方法研究了空间变异性对地震作用下真实 3D 边坡稳定性的影响。本研究中使用的随机响应面法 (VRSRSM) 与由于斜坡破坏中土壤参数的空间平均而产生的方差减少方法相结合，以减少由于在三维空间中实现空间相关的输入参数而导致的大量计算负荷。为了验证随机响应面法和方差缩减法对3维边坡可靠性的适用性，利用MCS和VRSRSM对无地震和地震条件下的简单3维边坡安全系数进行了分析。