Scholarlink

<p>This study presents the prediction of bearing capacity factors for multiple square shallow foundations in cohesive-frictional soils, utilizing finite element limit analysis (FELA), and machine learning (ML) techniques. The footings are considered to be of equal spacing <i>s</i>, and constant width<i> B</i>. Results from FELA, based on upper and lower bound theorems, were presented in dimensionless charts, showing the correlation between three bearing capacity factors (<i>N</i><sub><i>c</i></sub>, <i>N</i><sub><i>q</i></sub>, and <i>N</i><sub><i>γ</i></sub>), the angle of internal friction (<i>ϕ</i>), and the spacing ratio (<i>S/B</i>). ML techniques, namely ANN and XGBoost, were employed to estimate bearing capacity factors using <i>ϕ</i> and <i>S/B</i> as inputs. The developed models were assessed against FELA data through various metrics, with both ML models showing good agreement with FELA. Among the two models, XGBoost demonstrates slightly higher consistency with FELA data, with <i>R</i><sup>2</sup> values exceeding 99.9% across all datasets. Besides, a feature importance analysis identified the friction angle as the dominant parameter with permutation importance of more than 85% in the estimation of three bearing capacity factors.</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 Bearing Capacity Factors of Multiple Shallow Foundations Using Finite Element Limit Analysis and Machine Learning Approaches

  • 期刊论文

  • 发表日期：

    2025年2月26日

查看全文
  • Kittiphan Yoonirundorn,
  • Teerapong Senjuntichai,
  • Angsumalin Senjuntichai,
  • Suraparb Keawsawasvong

摘要

This study presents the prediction of bearing capacity factors for multiple square shallow foundations in cohesive-frictional soils, utilizing finite element limit analysis (FELA), and machine learning (ML) techniques. The footings are considered to be of equal spacing s, and constant width B. Results from FELA, based on upper and lower bound theorems, were presented in dimensionless charts, showing the correlation between three bearing capacity factors (Nc, Nq, and Nγ), the angle of internal friction (ϕ), and the spacing ratio (S/B). ML techniques, namely ANN and XGBoost, were employed to estimate bearing capacity factors using ϕ and S/B as inputs. The developed models were assessed against FELA data through various metrics, with both ML models showing good agreement with FELA. Among the two models, XGBoost demonstrates slightly higher consistency with FELA data, with R2 values exceeding 99.9% across all datasets. Besides, a feature importance analysis identified the friction angle as the dominant parameter with permutation importance of more than 85% in the estimation of three bearing capacity factors.