Reliability analysis approach provides a rational means to quantitatively evaluate the safety of geotechnical structures from a probabilistic perspective. However, it suffers from a known criticism of extensive computational requirements and poor efficiency, which hinders its application in the reliability analysis of earth dam slope stability. Until now, the effects of spatially variable soil properties on the earth dam slope reliability remain unclear. This calls for a novel method to perform reliability analysis of earth dam slope stability accounting for the spatial variability of soil properties. This paper develops an efficient extreme gradient boosting (XGBoost)-based reliability analysis approach for evaluating the earth dam slope failure probability. With the aid of the proposed approach, the failure probability of earth dam slope can be evaluated rationally and efficiently. The proposed approach is illustrated using a practical case adapted from Ashigong earth dam. Results show that the XGBoost-based reliability analysis approach is able to predict the earth dam slope failure probability with reasonable accuracy and efficiency. The coefficient of variations and scale of fluctuations of soil properties affect the earth dam slope failure probability significantly. Moreover, the earth dam slope failure probability is highly dependent on the selection of auto-correlation function (ACF), and the widely used single exponential ACF tends to provide an unconservative estimate in this study.

可靠性分析方法提供了一种从概率角度定量评估岩土结构安全性的合理方法。但是，它遭受了广泛的计算要求和效率低下的已知批评，这阻碍了其在土坝边坡稳定性可靠度分析中的应用。到目前为止，尚不清楚空间可变土壤特性对土坝坝坡可靠度的影响。这就需要一种新颖的方法来对土坝边坡稳定性进行可靠度分析，以解决土壤特性的空间变化问题。本文开发了一种基于有效的极端梯度提升（XGBoost）的可靠性分析方法，用于评估土坝边坡的破坏概率。借助建议的方法，可以合理有效地评估土坝边坡的破坏概率。所提出的方法是使用Ashigong土坝的实际案例进行说明的。结果表明，基于XGBoost的可靠性分析方法能够以合理的准确性和效率来预测土坝边坡的破坏概率。土性的变化系数和波动尺度对土坝的边坡破坏概率有显着影响。此外，土坝的边坡破坏概率高度依赖于自相关函数（ACF）的选择，在这项研究中，广泛使用的单指数ACF倾向于提供不保守的估计。结果表明，基于XGBoost的可靠性分析方法能够以合理的准确性和效率来预测土坝边坡的破坏概率。土性的变化系数和波动尺度对土坝的边坡破坏概率有显着影响。此外，土坝的边坡破坏概率高度依赖于自相关函数（ACF）的选择，在这项研究中，广泛使用的单指数ACF倾向于提供不保守的估计。结果表明，基于XGBoost的可靠性分析方法能够以合理的准确性和效率来预测土坝边坡的破坏概率。土性的变化系数和波动尺度对土坝的边坡破坏概率有显着影响。此外，土坝的边坡破坏概率高度依赖于自相关函数（ACF）的选择，在这项研究中，广泛使用的单指数ACF倾向于提供不保守的估计。