Scholarlink

<p>This study presents a hybrid computational intelligence framework for assessing the stability of square tunnels excavated in sloping rock masses following the Generalized Hoek–Brown failure criterion under surcharge loading, with the objective to develop a rapid and interpretable predictive tool for estimating the stability number (<i>N</i><sub><i>s</i></sub>). Adaptive Finite Element Limit Analysis (<i>AFELA</i>) was employed to simulate tunnel behaviour, and a database comprising 2700 numerical simulations was used by considering the effects of cover-depth ratio (<i>C/H</i>), slope inclination (<i>θ</i>), Geological Strength Index (<i>GSI</i>), material constant (<i>m</i><sub><i>i</i></sub>), and normalized uniaxial compressive strength (<i>σ</i><sub><i>ci</i></sub><i>/γH</i>). The dataset was integrated with tree-based machine learning models, including Classification and Regression Tree (CART), Chi-square Automatic Interaction Detector (CHAID), and Exhaustive CHAID (Ex-CHAID), for predicting <i>N</i><sub><i>s</i></sub>. Performance evaluation indicated that the CHAID model outperformed the CART and Ex-CHAID models, achieving the highest prediction accuracy with testing values of <i>R²</i> = 0.820 and RMSE = 3.755, while Taylor diagram analysis further confirmed its superior agreement with observed values. SHAP analysis revealed that <i>GSI</i> was the most influential parameter, followed by <i>m</i><sub><i>i</i></sub> and <i>C/H</i>, whereas <i>σ</i><sub><i>ci</i></sub><i>/γH</i> and <i>θ</i> exhibited comparatively lower influence on tunnel stability. The proposed <i>AFELA</i>–machine learning framework provides a computationally efficient, reliable, and interpretable approach for tunnel stability assessment in sloping rock mass 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

Tree Based Machine Learning Framework for Stability Evaluation of Square Tunnels in Sloping Rock

  • 期刊论文

  • 发表日期：

    2026年7月16日

查看全文
  • Anish Kumar,
  • Vinay Bhushan Chauhan,
  • Aayush Kumar,
  • Gurmesh Sihag,
  • Vikash Singh

摘要

This study presents a hybrid computational intelligence framework for assessing the stability of square tunnels excavated in sloping rock masses following the Generalized Hoek–Brown failure criterion under surcharge loading, with the objective to develop a rapid and interpretable predictive tool for estimating the stability number (Ns). Adaptive Finite Element Limit Analysis (AFELA) was employed to simulate tunnel behaviour, and a database comprising 2700 numerical simulations was used by considering the effects of cover-depth ratio (C/H), slope inclination (θ), Geological Strength Index (GSI), material constant (mi), and normalized uniaxial compressive strength (σci/γH). The dataset was integrated with tree-based machine learning models, including Classification and Regression Tree (CART), Chi-square Automatic Interaction Detector (CHAID), and Exhaustive CHAID (Ex-CHAID), for predicting Ns. Performance evaluation indicated that the CHAID model outperformed the CART and Ex-CHAID models, achieving the highest prediction accuracy with testing values of R² = 0.820 and RMSE = 3.755, while Taylor diagram analysis further confirmed its superior agreement with observed values. SHAP analysis revealed that GSI was the most influential parameter, followed by mi and C/H, whereas σci/γH and θ exhibited comparatively lower influence on tunnel stability. The proposed AFELA–machine learning framework provides a computationally efficient, reliable, and interpretable approach for tunnel stability assessment in sloping rock mass conditions.