基于自学习高斯过程回归模型的土石坝非饱和渗流稳定概率分析  

Probabilistic analysis of unsaturated seepage stability in embankment based on self-learning Gaussian Process Regression model

作　　者：朱国星 ZHU Guoxing(Rural Agricultural Development Authority of Qingshanhu District of Nanchang City in Jiangxi Province,Nanchang Jiangxi,330029,China)

机构地区：[1]江西省南昌市青山湖区农业农村局,江西南昌330029

出　　处：《江西水利科技》2025年第2期90-95,共6页Jiangxi Hydraulic Science & Technology

摘　　要：针对目前非饱和土石坝渗流稳定概率分析研究中的不足,文章提出了一种结合自学习策略与高斯过程回归模型(GPR)的土石坝渗流稳定概率分析方法。通过考虑土石坝两种材料参数的变异性,利用GPR模型构建土石坝随机输入变量与最小安全系数(FSmin)之间的关系,并采用主动学习策略确定最优训练样本数量,从而有效提高计算效率。通过某土石坝验证可知,该方法与传统蒙特卡洛模拟(MCS)相比,计算效率提升约125倍。该方法为考虑不确定性因素的土石坝非饱和渗流稳定性分析提供了一种高效而准确的工具。To address the current research limitations in the probabilistic stability analysis of unsaturatedembankments,this study proposes a novel probabilistic analysis approach for embankments stability,combining a self-learning strategy with the Gaussian Process Regression(GPR)model.By considering the variabilities of soil parameterswithin embankment,the GPR model was used to establish the relationship between the random input variables and theminimum safety factor(FSmin)of embankment.An active learning strategy was employed to determine the optimalnumber of training samples,thereby significantly enhancing the computational efficiency.Validation through anembankment case demonstrates that the proposed method achieves a computational efficiency approximately 125 timeshigher than traditional Monte Carlo Simulations(MCS)method.The proposed method provides an efficient and accuratetool for the probabilistic analysis of unsaturated embankments stability,accounting for uncertainty factors.

关 键 词：土石坝 非饱和 高斯过程回归模型 自学习策略 概率分析 

分 类 号：TV221.2[水利工程—水工结构工程]