Scholarlink

<p>This study addresses a key limitation of the traditional Wu–Waldron root-soil model (WWM): its reliance on a static, empirical k coefficient. We introduce a novel, data-driven framework to predict a dynamic k coefficient using Random Forest and LightGBM machine learning algorithms. The models were trained and validated on a comprehensive dataset from direct shear tests on unsaturated clayey silt bio-stabilized with <i>Nandina domestica</i>, incorporating moisture content (W), root weight density (RWD), depth, normal stress (NS), and strain. Both algorithms performed exceptionally well, with the Random Forest model achieving a mean R<sup>2</sup> of 0.987. A SHAP feature importance analysis consistently identified RWD and Strain as the most significant predictors of root reinforcement. This research provides a generalizable methodology to replace the static k coefficient, offering a more nuanced and accurate tool that significantly improves the reliability of slope stability assessments in eco-geotechnical applications.</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

Data-Driven Modeling of Additional Root-Soil Stress in Unsaturated Clayey Silt at Varying Strain Levels

  • 期刊论文

  • 发表日期：

    2025年9月24日

查看全文
  • Masoud Ebrahimi Derakhshan,
  • Jaber Mamaghanian,
  • Hamid Reza Razeghi

摘要

This study addresses a key limitation of the traditional Wu–Waldron root-soil model (WWM): its reliance on a static, empirical k coefficient. We introduce a novel, data-driven framework to predict a dynamic k coefficient using Random Forest and LightGBM machine learning algorithms. The models were trained and validated on a comprehensive dataset from direct shear tests on unsaturated clayey silt bio-stabilized with Nandina domestica, incorporating moisture content (W), root weight density (RWD), depth, normal stress (NS), and strain. Both algorithms performed exceptionally well, with the Random Forest model achieving a mean R2 of 0.987. A SHAP feature importance analysis consistently identified RWD and Strain as the most significant predictors of root reinforcement. This research provides a generalizable methodology to replace the static k coefficient, offering a more nuanced and accurate tool that significantly improves the reliability of slope stability assessments in eco-geotechnical applications.