Time-varying reliability analysis method for reservoir slopes based on ensembles of different machine learning algorithms

TIAN Mi; CHANG Hongjun; ZHONG Ruze; XIE Shiping; YANG Fan

doi:10.14188/j.1671-8844.2025.0155

您当前的位置：

首页 >

文章列表页 >

Time-varying reliability analysis method for reservoir slopes based on ensembles of different machine learning algorithms

更新时间：2026-10-15

- Time-varying reliability analysis method for reservoir slopes based on ensembles of different machine learning algorithms
- Engineering Journal of Wuhan University Pages: 1-15(2026)
- 作者机构：
  
  1.湖北工业大学土木建筑与环境学院，湖北武汉 430068
  2.河湖健康智慧感知与生态修复教育部重点实验室，武汉 430068
  3.长沙矿山研究院有限责任公司，湖南长沙 410012
  4.长江水利水电开发集团（湖北）有限公司，湖北武汉 430014
- 作者简介：
- 基金信息：
- DOI：10.14188/j.1671-8844.2025.0155
  CLC： TU43
- Online First：15 October 2026，
- 稿件说明：
移动端阅览
TIAN Mi, CHANG Hongjun, ZHONG Ruze, et al. Time-varying reliability analysis method for reservoir slopes based on ensembles of different machine learning algorithms[J/OL]. Engineering Journal of Wuhan University, 2026, 1-15.
DOI：

TIAN Mi, CHANG Hongjun, ZHONG Ruze, et al. Time-varying reliability analysis method for reservoir slopes based on ensembles of different machine learning algorithms[J/OL]. Engineering Journal of Wuhan University, 2026, 1-15. DOI： 10.14188/j.1671-8844.2025.0155.

摘要

基于机器学习代理模型的边坡可靠度分析可有效减少确定性分析的计算次数而降低计算量、提高效率。目前采用机器学习分析边坡可靠度通常基于不同机器学习代理模型的预测性能选择最优的单个机器学习。然而，选择单一机器学习无法充分利用其他算法的优点，且存在泛化能力不足的问题。为此，本文提出了一种融合不同机器学习的考虑岩土体参数空间变异性的库岸边坡时变可靠度分析方法。首先，采用拉丁超立方抽样技术产生输入参数（包含岩土体抗剪强度参数、水位降速）的随机样本；基于支持向量机（SVM）和多元自适应回归（MARS）算法分别构建库岸边坡稳定性分析等效代理模型获取边坡安全系数确定性预测；然后基于贝叶斯模型平均提出融合SVM和MARS算法的库岸边坡稳定性分析集合代理模型。最后，通过蒙特卡洛模拟方法计算库岸边坡时变可靠度，给出边坡安全系数不确定性预测区间，并对比单一机器学习代理模型与集合代理模型计算库岸边坡稳定性的准确性。本文以三峡库区重庆段某库岸边坡为例，阐明了融合不同机器学习的库岸边坡时变可靠度分析方法的有效性。结果表明：相比于单一机器学习代理模型，融合SVM和MARS算法的库岸边坡稳定性分析集合概率模型预测的边坡安全系数更接近于数值模型计算值，其均方误差和平均绝对误差平均降低3.6%和15.6%。所构建的边坡稳定性集合代理模型不仅能准确高效地计算边坡时变失效概率，而且能合理考虑代理模型的预测不确定性，可为库岸边坡可靠度问题计算提供有效的分析方法。

Abstract

Machine learning (ML) based surrogate models can effectively reduce the number of deterministic simulation analyses

reduce computational costs and improve computational efficiency. At present

ML based slope reliability analysis usually selects an optimal single model among alternative surrogate models based on their predictive performance. However

selecting a single ML model cannot completely employ the advantages of other algorithms and may suffer from limited generalization capability. To address this issue

this study proposes a time-varying reliability analysis method for reservoir slopes by integrating different machine learning models while considering the spatial variability of geotechnical parameters. Firstly

Latin hypercube sampling (LHS) is used to generate random samples of input parameters (including shear strength parameters and the water level drawdown). Then

surrogate models for slope stability analysis are constructed based on support vector machines (SVM) and multivariate adaptive regression splines (MARS)

respectively. Subsequently

deterministic predictions of slope safety factors by SVM and MARS are integrated by the Bayesian model averaging (BMA) approach to generate an ensemble surrogate model for slope stability analysis. Finally

the time-varying reliability of reservoir slope is calculated by Monte Carlo simulation method and the uncertainty prediction intervals for slope safety factors are also provided. The prediction accuracy of single ML models and the ensemble model are also investigated. A slope in Chongqing section of the Three Gorges reservoir area is taken as an example to illustrate the proposed method. Results show that

compared with single ML models

the slope safety factors estimated from the BMA ensemble model are closer to the numerical simulation results. Furthermore

the mean squared error (MSE) and mean absolute error (MAE) of the ensemble model are reduced by an average of 3.6% and 15.6%

respectively. The proposed BMA ensemble model can not only accurately estimate the time-varying failure probability of slopes

but also reasonably account for the predictive uncertainty of surrogate models. This approach provides an effective analytical method for addressing the reliability of reservoir slopes.

关键词

Keywords

references

苏爱军 . 湖北省三峡库区滑坡防治地质勘察与治理工程技术规定 [M ] . 武汉 : 中国地质大学出版社 , 2003 .

Su A J . Technical regulation of geological investigation and engineering design for landside in Three Gorges Reservoir, Hubei Province [M ] . Wuhan : China university of Geosciences press , 2003 . (in Chinese with English abstract)

李长冬 , 龙晶晶 , 姜茜慧 , 等 . 水库滑坡成因机制研究进展与展望 [J ] . 地质科技通报 , 2020 , 39 ( 1 ): 67 - 77 .

Li C D , Long J J , Jiang X H , et al . Advance and prospect of formation mechanism for reservoir landslide [J ] . Bulletin of Geological Science and Technology , 2020 , 39 ( 1 ): 67 - 77 . (in Chinese with English abstract)

仉文岗 , 王尉 , 高学成 . 库区水位下降对库岸边坡稳定性的影响 [J ] . 武汉大学学报（工学版） , 2019 , 52 ( 1 ): 21 - 26 .

Zhang W G , Wang W , Gao X C . Influences of reservoirs water level drawdown on slope stability [J ] . Engineering Journal of Wuhan University , 2019 , 52 ( 1 ): 21 - 26 . (in Chinese with English abstract)

刘新喜 , 夏元友 , 张显书 , 等 . 库水位下降对滑坡稳定性的影响 [J ] . 岩石力学与工程学报 , 2005 , ( 08 ): 1439 - 1444 .

Liu X X , Xia Y Y , Zhang X S , et al . Effects of drawdown of reservoir water level on landslide stability . [J ] . Chinese Journal of Rock Mechanics and Engineering , 2005 , ( 08 ): 1439 - 1444 . (in Chinese with English abstract)

廖红建 , 盛谦 , 高石夯 , 等 . 库水位下降对滑坡体稳定性的影响 [J ] . 岩石力学与工程学报 , 2005 , ( 19 ): 56 - 60 .

Liao H J , Sheng Q , Gao S H , et al . Influence of drawdown of reservoir water level on landslide stability . [J ] . Chinese Journal of Rock Mechanics and Engineering , 2005 , ( 19 ): 56 - 60 . (in Chinese with English abstract)

李典庆 , 蒋水华 , 周创兵 , 等 . 考虑参数空间变异性的边坡可靠度分析非侵入式随机有限元法 [J ] . 岩土工程学报 , 2013 , 35 ( 08 ): 1413 - 1422 .

Li D Q , Jiang S H , Zhou C B , et al . Reliability analysis of slopes considering spatial variability of soil parameters using non-intrusive stochastic finite element method . [J ] . Chinese Journal of Geotechnical Engineering , 2013 , 35 ( 08 ): 1413 - 1422 . (in Chinese with English abstract)

武梦婷 , 陈秋松 , 齐冲冲 . 基于机器学习的边坡安全稳定性评价及防护措施 [J ] . 工程科学学报 , 2022 , 44 ( 02 ): 180 - 188 .

Wu M T , Chen Q S , Qi C C . Slope safety, stability evaluation, and protective measures based on machine learning [J ] . Chinese Journal of Engineering , 2022 , 44 ( 02 ): 180 - 188 . (in Chinese with English abstract)

姬建 , 姜振 , 殷鑫 , 等 . 边坡随机场数字图像特征CNN深度学习及可靠度分析 [J ] . 岩土工程学报 , 2022 , 44 ( 08 ): 1463 - 1473 .

Ji J , Jiang Z , Yin X , et al . Slope reliability analysis based on deep learning of digital images of random fields using CNN [J ] . Chinese Journal of Geotechnical Engineering , 2022 , 44 ( 08 ): 1463 - 1473 . (in Chinese with English abstract)

Wang L , Wu C , Tang L , et al . Efficient reliability analysis of earth dam slope stability using extreme gradient boosting method [J ] . Acta Geotechnica , 2020 , 15 : 3135 - 3150 .

Liu L L , Zhang S H , Cheng Y M , et al . Advanced reliability analysis of slopes in spatially variable soils using multivariate adaptive regression splines. Geosci Front 10 (2): 671–682 [J ] . 2019 .

张天龙 , 曾鹏 , 李天斌 , 等 . 基于主动学习径向基函数的边坡系统可靠度分析 [J ] . 岩土力学 , 2020 , 41 ( 09 ): 3098 - 3108 .

Zhang T L , Zeng P , Li T B , et al . System reliability analyses of slopes based on active-learning radial basis function [J ] . Rock and Soil Mechanics , 2020 , 41 ( 09 ): 3098 - 3108 . (in Chinese with English abstract)

Zhang W G , Wu C Z , Tang L B , et al . Efficient time-variant reliability analysis of Bazimen landslide in the Three Gorges Reservoir Area using XGBoost and LightGBM algorithms [J ] . Gondwana Research , 2023 , 123 : 41 - 53 .

Mckay M D , Beckman R J , Conover W J . A comparison of three methods for selecting values of input variables in the analysis of output from a computer code [J ] . Technometrics , 1979 , 21 ( 2 ): 239 - 245 .

蒋水华 , 李典庆 , 周创兵 . 基于拉丁超立方抽样的边坡可靠度分析非侵入式随机有限元法 [J ] . 岩土工程学报 , 2013 , 35 ( S2 ): 70 - 76 .

Jiang S H , Li D Q , Zhou C B . Non-intrusive stochastic finite element method for slope reliability analysis based on Latin hypercube sampling [J ] . Chinese Journal of Geotechnical Engineering , 2013 , 35 ( S2 ): 70 - 76 . (in Chinese with English abstract)

吴振君 , 王水林 , 葛修润 . LHS方法在边坡可靠度分析中的应用 [J ] . 岩土力学 , 2010 , 31 ( 4 ): 1047 - 1054 .

Wu Z J , Wang S L , Ge X R . Application of Latin hypercube sampling technique to slope reliability analysis [J ] . Rock and Soil Mechanics , 2010 , 31 ( 4 ): 1047 - 1054 . (in Chinese with English abstract)

Fredlund D G , Rahardjo H , Fredlund M D . Unsaturated soil mechanics in engineering practice [M ] . John Wiley & Sons , 2012 .

Fredlund D G , Xing A . Equations for the soil-water characteristic curve [J ] . Canadian geotechnical journal , 1994 , 31 ( 4 ): 521 - 532 .

Chen Z Y , Morgenstern N R . Extensions to the generalized method of slices for stability analysis [J ] . Canadian geotechnical journal , 1983 , 20 ( 1 ): 104 - 119 .

Cortes C , Vapnik V . Support-vector networks [J ] . Machine Learning , 1995 , 20 : 273 - 297 .

Friedman J H . Multivariate adaptive regression splines [J ] . Annals of Statistics , 1991 , 19 : 1 - 67 .

仉文岗 , 洪利 , 黎泳钦 . 基于多元自适应回归样条的高维岩土工程问题分析 [J ] . 河海大学学报( 自然科学版) , 2019 , 47 ( 4 ): 359 - 365 .

Zhang W G , Hong L , Li Y Q . Analysis of multi-dimensional geotechnical engineering problems based on

multivariate adaptive regression splines [J ] . Journal of Hohai University( Natural Sciences) , 2019 , 47 ( 4 ): 359 - 365 .

董磊华 , 熊立华 , 万民 . 基于贝叶斯模型加权平均方法的水文模型不确定性分析 [J ] . 水利学报 , 2011 , 42 ( 9 ): 1065 - 1074 .

Dong L H , Xiong L H , Wan M . Uncertainty analysis of hydrological modeling using the Bayesian Model Averaging Method [J ] . Shuili Xuebao , 2011 , 42 ( 9 ): 1065 - 1074 . (in Chinese with English abstract)

田密 , 熊自民 , 李丽华 . 基于多机器学习融合的泥石流冲出距离贝叶斯集合概率预测方法 [J ] . 武汉大学学报(工学版): 1 - 18 , 2024 . (网络首发)

Tian M , Xiong Z M , Li L H . Bayesian ensemble probabilistic prediction method for debris-flow inundated area based on ensembles of multiple machine learning algorithms [J ] . Engineering Journal of Wuhan University: 1 - 18 , 2024 . (in Chinese with English abstract)

仉文岗 , 顾鑫 , 刘汉龙 , 等 . 基于贝叶斯更新的非饱和土坡参数概率反演及变形预测 [J ] . 岩土力学 , 2022 , 43 ( 04 ): 1112 - 1122 .

Zhang W G , Gu X , Liu H L , et al . Probabilistic back analysis of soil parameters and displacement prediction of unsaturated slopes using Bayesian updating [J ] . Rock and Soil Mechanics , 2022 , 43 ( 04 ): 1112 - 1122 . (in Chinese with English abstract)

Vrugt J A , Robinson B A . Treatment of uncertainty using ensemble methods: Comparison of sequential data assimilation and Bayesian model averaging [J ] . Water Resources Research , 2007 , 43 ( 1 ).

Vrugt J A , Diks C G H , Clark M P . Ensemble Bayesian model averaging using Markov chain Monte Carlo sampling [J ] . Environmental fluid mechanics , 2008 , 8 : 579 - 595 .

Vrugt J A , Ter Braak C J F , Clark M P , et al . Treatment of input uncertainty in hydrologic modeling: Doing hydrology backward with Markov chain Monte Carlo simulation [J ] . Water Resources Research , 2008 , 44 ( 12 ).

Vrugt J A . Markov chain Monte Carlo simulation using the DREAM software package: Theory, concepts, and MATLAB implementation [J ] . Environmental Modelling & Software , 2016 , 75 : 273 - 316 .

Tan X H , Hu N , Li D , et al . Time-variant reliability analysis of unsaturated soil slopes under rainfall [J ] . Geotechnical & Geological Engineering , 2013 , 31 : 319 - 327 .

Wu Y P , Miao F S , Li L W , et al . Time-varying reliability analysis of Huangtupo Riverside No.2 Landslide in the Three Gorges Reservoir based on water-soil coupling [J ] . Engineering Geology , 2017 , 226 ( 2 ): 267 – 276 .

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Time‑varying reliability analysis method of reservoir slopes considering the spatial variability of rock and soil parameters and the uncertainty of water level drawdown

Machine learning based optimization study of charging pile ventilation grille structure

Water demand forecasting based on interpretable machine learning models and Monte Carlo method

Digital evaluation of the transmission tower operation status based on stress monitoring

A Bayesian ensemble probabilistic prediction method for the debris⁃flow runout distance based on fusion of multiple machine learning algorithms

Related Author

CHANG Hongjun

YANG Fan

WU Shijing

CHENG Han

ZHANG Nan

YUE Yanan

WANG Huijiu

YANG Limin

Related Institution

Changjiang Water Resources and Hydropower Development Group Co.， Ltd.

Bull Industrial Research Institute of New Energy， Wuhan University

New Energy Division， Bull Group

School of Power and Mechanical Engineering， Wuhan University

State Key Laboratory of Water Resources Engineering and Management， Wuhan University

AI问答

⁰