Scholarlink

<p>The inherent spatial variability of geomaterials, resulting from complex sedimentation processes, introduces significant uncertainty in assessing slope stability. This variability strongly affects both the factors of safety (<i>F</i><sub><i>s</i></sub>) and the configuration of critical slip surfaces (CSS), often leading to multiple potential failure modes. To address these challenges, this study proposes a novel Intelligent Stochastic Meshless Numerical Manifold Method (ISMNMM) framework, which integrates the Complex Variable Meshfree Numerical Manifold Method (CVMNMM), Gaussian random fields, and a Self-Organizing Map Neural Network (SOMNN). The SOMNN is innovatively introduced into the strength reduction method (SRM) as a new convergence criterion and as an intelligent tool for the automatic extraction of smooth CSS, effectively overcoming the “over-reduction” issue and improving the reliability of <i>F</i><sub><i>s</i></sub> and CSS estimation. Two benchmark slope examples demonstrate that the proposed method can accurately identify both circular and noncircular CSS. A stochastic analysis of an engineering slope further reveals three dominant failure modes—shallow, intermediate, and deep—among which the intermediate failure accounts for approximately 64% of occurrences on average. The mean difference in Fs among the three failure modes compared with deterministic results was small (2.12% on average). However, the mean difference in the sliding mass volume reached up to 42.66%, highlighting the necessity of analyzing the stochastic distribution of CSS. The cross-correlation coefficient of random fields has a limited effect on the proportion of failure modes but notably influences the scale of deep-seated CSS. Based on these findings, medium-length anchors (10 ~ 20&#xa0;m) are recommended as the primary reinforcement measure, complemented by longer anchors in locally deep failure zones. Overall, the ISMNMM provides an efficient and intelligent framework for quantifying the stochastic behavior of slopes under spatial variability and offers valuable guidance for sustainable slope stabilization design and hazard prevention in complex geological conditions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Stochastic analysis of slope stability based on a novel complex variable meshfree numerical manifold method and self-organizing map neural network

  • 期刊论文

  • 发表日期：

    2025年12月23日

查看全文
  • Xitailang Cao,
  • Shan Lin,
  • Hongwei Guo,
  • Ruiliang Sun,
  • Quanke Hu,
  • Hong Zheng,
  • Zhijun Liu

摘要

The inherent spatial variability of geomaterials, resulting from complex sedimentation processes, introduces significant uncertainty in assessing slope stability. This variability strongly affects both the factors of safety (Fs) and the configuration of critical slip surfaces (CSS), often leading to multiple potential failure modes. To address these challenges, this study proposes a novel Intelligent Stochastic Meshless Numerical Manifold Method (ISMNMM) framework, which integrates the Complex Variable Meshfree Numerical Manifold Method (CVMNMM), Gaussian random fields, and a Self-Organizing Map Neural Network (SOMNN). The SOMNN is innovatively introduced into the strength reduction method (SRM) as a new convergence criterion and as an intelligent tool for the automatic extraction of smooth CSS, effectively overcoming the “over-reduction” issue and improving the reliability of Fs and CSS estimation. Two benchmark slope examples demonstrate that the proposed method can accurately identify both circular and noncircular CSS. A stochastic analysis of an engineering slope further reveals three dominant failure modes—shallow, intermediate, and deep—among which the intermediate failure accounts for approximately 64% of occurrences on average. The mean difference in Fs among the three failure modes compared with deterministic results was small (2.12% on average). However, the mean difference in the sliding mass volume reached up to 42.66%, highlighting the necessity of analyzing the stochastic distribution of CSS. The cross-correlation coefficient of random fields has a limited effect on the proportion of failure modes but notably influences the scale of deep-seated CSS. Based on these findings, medium-length anchors (10 ~ 20 m) are recommended as the primary reinforcement measure, complemented by longer anchors in locally deep failure zones. Overall, the ISMNMM provides an efficient and intelligent framework for quantifying the stochastic behavior of slopes under spatial variability and offers valuable guidance for sustainable slope stabilization design and hazard prevention in complex geological conditions.