Scholarlink

<p>This paper presented a node-based smoothed radial point interpolation method with linear strain fields (NS-RPIM) for stability analysis of dual horseshoe tunnels in cohesive-frictional soils subjected to surcharge loadings. Using the upper bound theorem, dual horseshoe-shaped tunnels were modelled under plane strain conditions, and the soil was described as a Mohr-Coulomb material obeying an associated flow rule. The study investigated the variation in stability factors <i>N</i> = <i>σ</i><sub><i>s</i></sub><i>/c</i> with respect to the horizontal spacing ratio <i>S/B</i>, the tunnel cover depth ratio <i>H/B</i>, the height-to-width ratio <i>h/B</i>, the soil weight parameter <i>γB/c</i>, and the friction angle <i>φ</i>. A dataset comprising 2205 data points obtained from NS-RPIM results was further utilized to develop predictive machine learning models. The Extreme Gradient Boosting (XGBoost) model achieved excellent predictive accuracy (R² ≈ 0.999, RMSE &lt; 1.0 on the test set). SHapley Additive exPlanations (SHAP) and Partial Dependence Plots (PDP) analysis confirmed the dominant influence of the friction angle <i>φ</i>, and the spacing ratio <i>S/B</i>. The hybrid approach, which integrated the precision of NS-RPIM with the XGBoost model, provided a robust tool for preliminary tunnel design. The results were presented in the form of design tables and charts, offering valuable insights for geotechnical engineers. The proposed framework demonstrated scalability for broader geotechnical applications, enhancing both accuracy and computational efficiency.</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

A Hybrid of Node-Based Smoothed Radial Point Interpolation Method with Extreme Gradient Boosting for Stability Prediction of Dual Horseshoe Tunnels

  • 期刊论文

  • 发表日期：

    2026年3月20日

查看全文
  • L. Nguyen-Son,
  • T. Thai-Phuong,
  • T. Vo-Minh

摘要

This paper presented a node-based smoothed radial point interpolation method with linear strain fields (NS-RPIM) for stability analysis of dual horseshoe tunnels in cohesive-frictional soils subjected to surcharge loadings. Using the upper bound theorem, dual horseshoe-shaped tunnels were modelled under plane strain conditions, and the soil was described as a Mohr-Coulomb material obeying an associated flow rule. The study investigated the variation in stability factors N = σs/c with respect to the horizontal spacing ratio S/B, the tunnel cover depth ratio H/B, the height-to-width ratio h/B, the soil weight parameter γB/c, and the friction angle φ. A dataset comprising 2205 data points obtained from NS-RPIM results was further utilized to develop predictive machine learning models. The Extreme Gradient Boosting (XGBoost) model achieved excellent predictive accuracy (R² ≈ 0.999, RMSE < 1.0 on the test set). SHapley Additive exPlanations (SHAP) and Partial Dependence Plots (PDP) analysis confirmed the dominant influence of the friction angle φ, and the spacing ratio S/B. The hybrid approach, which integrated the precision of NS-RPIM with the XGBoost model, provided a robust tool for preliminary tunnel design. The results were presented in the form of design tables and charts, offering valuable insights for geotechnical engineers. The proposed framework demonstrated scalability for broader geotechnical applications, enhancing both accuracy and computational efficiency.