Scholarlink

<p>This study proposes a machine learning based predictive framework for estimating the stability number of circular tunnels constructed in sloping rock masses under surcharge loading. A comprehensive dataset generated through Adaptive Finite Element Limit Analysis simulations was used to train and evaluate multiple regression models, including Gradient Boosting Machine (GBM), Random Forest (RDF), Support Vector Machine (SVM) with radial basis function (RBF), linear, and polynomial kernels, and Linear Regression. A comparative assessment based on statistical error metrics, Taylor diagrams, and absolute error distribution analysis clearly demonstrated the superiority of nonlinear models over linear approaches. The GBM model achieved the highest predictive performance, with <i>R</i><sup><i>2</i></sup> = 1.000 and RMSE = 0.096 during training, and <i>R</i><sup><i>2</i></sup> = 0.990 and RMSE = 1.195 during testing, indicating strong generalization capability. The SVM-RBF model ranked second (<i>R</i><sup><i>2</i></sup> = 0.969 during testing), whereas linear-based models exhibited significantly lower accuracy. Interpretation of variable importance, sensitivity, and SHAP analyses consistently identified the Geological Strength Index and material constant as the primary predictors of tunnel stability. The proposed framework offers a computationally efficient and reliable tool for rapid stability assessment of tunnels in complex sloping rock environments.</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

Machine Learning Framework for Stability Prediction of Circular Tunnel in Sloping Rock Mass under Surcharge Loading

  • 期刊论文

  • 发表日期：

    2026年4月24日

查看全文
  • Anish Kumar,
  • Vinay Bhushan Chauhan,
  • Aayush Kumar

摘要

This study proposes a machine learning based predictive framework for estimating the stability number of circular tunnels constructed in sloping rock masses under surcharge loading. A comprehensive dataset generated through Adaptive Finite Element Limit Analysis simulations was used to train and evaluate multiple regression models, including Gradient Boosting Machine (GBM), Random Forest (RDF), Support Vector Machine (SVM) with radial basis function (RBF), linear, and polynomial kernels, and Linear Regression. A comparative assessment based on statistical error metrics, Taylor diagrams, and absolute error distribution analysis clearly demonstrated the superiority of nonlinear models over linear approaches. The GBM model achieved the highest predictive performance, with R2 = 1.000 and RMSE = 0.096 during training, and R2 = 0.990 and RMSE = 1.195 during testing, indicating strong generalization capability. The SVM-RBF model ranked second (R2 = 0.969 during testing), whereas linear-based models exhibited significantly lower accuracy. Interpretation of variable importance, sensitivity, and SHAP analyses consistently identified the Geological Strength Index and material constant as the primary predictors of tunnel stability. The proposed framework offers a computationally efficient and reliable tool for rapid stability assessment of tunnels in complex sloping rock environments.