Scholarlink

<p>Machine learning (ML) approaches have been widely employed for identifying engineering properties, stability analysis and ground improvement studies. However, there remains a lack of reliable forecasting methods for evaluating the seismic stability of soil structure using ML algorithms. Soil nailing is a reinforcement method employed to enhance the stability of soil mass under static and dynamic conditions. This study presents a method for assessing the stability of the nailed soil using pseudo-static and pseudo-dynamic approaches in limit-equilibrium framework. A realistic composite failure was considered and the influence of soil, geometric and seismic properties were studied. The results show an increase in the factor of safety (FOS) of approximately 25–35% compared to wedge-based pseudo-dynamic analyses and about 15–20% over simplified mechanisms, demonstrating that the difference is not only conceptual but also design-relevant. The results obtained under pseudo-static and pseudo-dynamic conditions indicate a 26% increase in the estimated stability. This comparative study exhibits the efficiency of pseudo-dynamic method over pseudo-static. For prescribed FOS value, improved stability estimates results in optimization of reinforcement parameters. This suggests that traditional assumption may lead to conservative designs, whereas the current approach provides a rational basis for more efficient and potentially economic solutions. This study aims to develop ML algorithms, namely Linear Regression, Stochastic Gradient Descent Regression, Support Vector Regression, Artificial Neural Network, and Random Forest to predict the factor of safety of the nailed soil mass. The ML algorithm demonstrated superior performance through repetitive cross-validation and hyperparameter tuning. This research provides useful recommendations for future seismic stability analysis of nailed soil mass.</p> Graphical abstract <p></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-based prediction of seismic stability of soil-nailed excavations with composite failure surfaces

  • 期刊论文

  • 发表日期：

    2026年6月2日

查看全文
  • Surya Muthukumar,
  • Dhanya Sathyan,
  • B. Premjith,
  • Sanjay Kumar Shukla,
  • Sreevalsa Kolathayar

摘要

Machine learning (ML) approaches have been widely employed for identifying engineering properties, stability analysis and ground improvement studies. However, there remains a lack of reliable forecasting methods for evaluating the seismic stability of soil structure using ML algorithms. Soil nailing is a reinforcement method employed to enhance the stability of soil mass under static and dynamic conditions. This study presents a method for assessing the stability of the nailed soil using pseudo-static and pseudo-dynamic approaches in limit-equilibrium framework. A realistic composite failure was considered and the influence of soil, geometric and seismic properties were studied. The results show an increase in the factor of safety (FOS) of approximately 25–35% compared to wedge-based pseudo-dynamic analyses and about 15–20% over simplified mechanisms, demonstrating that the difference is not only conceptual but also design-relevant. The results obtained under pseudo-static and pseudo-dynamic conditions indicate a 26% increase in the estimated stability. This comparative study exhibits the efficiency of pseudo-dynamic method over pseudo-static. For prescribed FOS value, improved stability estimates results in optimization of reinforcement parameters. This suggests that traditional assumption may lead to conservative designs, whereas the current approach provides a rational basis for more efficient and potentially economic solutions. This study aims to develop ML algorithms, namely Linear Regression, Stochastic Gradient Descent Regression, Support Vector Regression, Artificial Neural Network, and Random Forest to predict the factor of safety of the nailed soil mass. The ML algorithm demonstrated superior performance through repetitive cross-validation and hyperparameter tuning. This research provides useful recommendations for future seismic stability analysis of nailed soil mass.

Graphical abstract