Scholarlink

<p>This study investigates the uplift capacity of circular anchors embedded in frictional-cohesive soils under surcharge. The analysis focuses on three critical stability factors <i>F</i><sub><i>c</i></sub>, <i>F</i><sub><i>q</i></sub>, and <i>F</i><sub><i>γ</i></sub> using Terzaghi’s principle of superposition to evaluate ultimate bearing capacity. These factors are influenced by the soil’s internal friction angle, the geometric ratio of anchor depth to diameter, and the interface roughness between the anchor and soil. Three predictive models for these stability factors are developed using advanced computational methods, including finite element limit analysis (FELA) with adaptive meshing and Kolmogorov-Arnold Networks (KAN). This research is the first to apply KAN in anchor behavior studies, demonstrating its enhanced ability to model complex data relationships compared to artificial neural networks (ANN). Additionally, a closed-form solution for stability factors is derived through KAN, providing an efficient method for predicting bearing capacity. The optimized models exhibit high coefficient of determination (R²) values and low root mean square errors (RMSE) for training and testing datasets. Sensitivity analysis validates the robustness of the proposed models. These findings advance the understanding of circular anchors’ bearing capacity in frictional-cohesive soils, offering practical design insights for various soil conditions.</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

Predicting the uplift capacity of circular anchors in frictional-cohesive soils using Kolmogorov-Arnold networks

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年4月25日

查看全文
  • Tran Vu-Hoang,
  • Tan Nguyen,
  • Jim Shiau,
  • Hung-Thinh Pham-Tran,
  • Trung Nguyen-Thoi

摘要

This study investigates the uplift capacity of circular anchors embedded in frictional-cohesive soils under surcharge. The analysis focuses on three critical stability factors Fc, Fq, and Fγ using Terzaghi’s principle of superposition to evaluate ultimate bearing capacity. These factors are influenced by the soil’s internal friction angle, the geometric ratio of anchor depth to diameter, and the interface roughness between the anchor and soil. Three predictive models for these stability factors are developed using advanced computational methods, including finite element limit analysis (FELA) with adaptive meshing and Kolmogorov-Arnold Networks (KAN). This research is the first to apply KAN in anchor behavior studies, demonstrating its enhanced ability to model complex data relationships compared to artificial neural networks (ANN). Additionally, a closed-form solution for stability factors is derived through KAN, providing an efficient method for predicting bearing capacity. The optimized models exhibit high coefficient of determination (R²) values and low root mean square errors (RMSE) for training and testing datasets. Sensitivity analysis validates the robustness of the proposed models. These findings advance the understanding of circular anchors’ bearing capacity in frictional-cohesive soils, offering practical design insights for various soil conditions.