Scholarlink

<p>Accurate prediction of soil slope stability plays an important role in geotechnical engineering and helps reduce human casualties and financial losses. This study presents an efficient framework for the probabilistic analysis and prediction of soil slope stability using Artificial Neural Networks (ANN) and Subset Simulation (SS). A MATLAB-based tool is developed to perform finite element simulations and generate synthetic datasets for ANN training. The trained model is then used within the SS method to estimate slope stability reliability metrics. Validation through case studies shows that the obtained ANN model predicts the Factor of Safety (FS) with an average error of 1.8%. Compared to traditional random variable and random field approaches, the proposed ANN–SS method achieves similar accuracy, with differences in the reliability index (β) limited to 11% and 14%, respectively. A case study of the Shiraz–Siyakh roadside slope further demonstrates that the proposed approach can effectively assess design performance, revealing that FS alone may not adequately reflect true slope stability 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

Probabilistic Stability Analysis of Slope Using Artificial Neural Network and Subset Simulation

  • 期刊论文

  • 发表日期：

    2025年9月16日

查看全文
  • P. Karimi,
  • A. Gholampour

摘要

Accurate prediction of soil slope stability plays an important role in geotechnical engineering and helps reduce human casualties and financial losses. This study presents an efficient framework for the probabilistic analysis and prediction of soil slope stability using Artificial Neural Networks (ANN) and Subset Simulation (SS). A MATLAB-based tool is developed to perform finite element simulations and generate synthetic datasets for ANN training. The trained model is then used within the SS method to estimate slope stability reliability metrics. Validation through case studies shows that the obtained ANN model predicts the Factor of Safety (FS) with an average error of 1.8%. Compared to traditional random variable and random field approaches, the proposed ANN–SS method achieves similar accuracy, with differences in the reliability index (β) limited to 11% and 14%, respectively. A case study of the Shiraz–Siyakh roadside slope further demonstrates that the proposed approach can effectively assess design performance, revealing that FS alone may not adequately reflect true slope stability conditions.