Scholarlink

<p>In this study, eight steep road-cut slopes in a seismically sensitive mountainous region were analysed under static and pseudo-static conditions using a combined analytical and data-driven framework. Rock mass conditions were characterised through rock mass rating (RMR), slope mass rating (SMR), and geological strength index (GSI) to establish field-based validation. Geotechnical parameters, including slope height, slope angle, GSI, and uniaxial compressive strength (UCS), were determined from field surveys, laboratory testing, and empirical correlations. The limit equilibrium method (LEM) incorporating the generalized Hoek–Brown (GHB) failure criterion indicated that five slopes were unstable under static conditions, whereas all slopes became unstable under pseudo-static conditions, reflecting the influence of steep geometries and weak rock masses. To enhance predictive capability, an automated machine learning framework was developed and trained on 350 slope cases, comprising 50 cases compiled from published literature and 300 synthetically generated through numerical modelling, and tested on the slopes under investigation. Model evaluation was carried out using the coefficient of determination, root mean squared error, mean absolute percentage error, and spearman’s correlation coefficient. The ensemble model reproduced the computed safety factors with high agreement and consistently identified marginally stable and critical slopes. Feature importance analysis showed that the GSI and UCS exerted greater influence than slope geometry.</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

Static and Pseudo-static Stability of Road-cuts by Merging Limit Equilibrium and Auto-ML Frameworks

  • 期刊论文

  • 发表日期：

    2025年11月21日

查看全文
  • Virat Singh Chauhan,
  • Md Rehan Sadique,
  • Mohd Masroor Alam,
  • Mohd Ahmadullah Farooqi

摘要

In this study, eight steep road-cut slopes in a seismically sensitive mountainous region were analysed under static and pseudo-static conditions using a combined analytical and data-driven framework. Rock mass conditions were characterised through rock mass rating (RMR), slope mass rating (SMR), and geological strength index (GSI) to establish field-based validation. Geotechnical parameters, including slope height, slope angle, GSI, and uniaxial compressive strength (UCS), were determined from field surveys, laboratory testing, and empirical correlations. The limit equilibrium method (LEM) incorporating the generalized Hoek–Brown (GHB) failure criterion indicated that five slopes were unstable under static conditions, whereas all slopes became unstable under pseudo-static conditions, reflecting the influence of steep geometries and weak rock masses. To enhance predictive capability, an automated machine learning framework was developed and trained on 350 slope cases, comprising 50 cases compiled from published literature and 300 synthetically generated through numerical modelling, and tested on the slopes under investigation. Model evaluation was carried out using the coefficient of determination, root mean squared error, mean absolute percentage error, and spearman’s correlation coefficient. The ensemble model reproduced the computed safety factors with high agreement and consistently identified marginally stable and critical slopes. Feature importance analysis showed that the GSI and UCS exerted greater influence than slope geometry.