Scholarlink

<p>Accurate estimation of undrained bearing capacity near slopes was crucial for geotechnical stability, particularly in anisotropic clay where the strength varied by direction. This study employed finite element limit analysis (FELA), based on the AUS failure criterion under plane strain conditions, to generate reference bearing capacity values. Normalized input variables were analyzed: undrained shear strength ratio (<i>s</i><sub><i>uc</i></sub><i>/γB</i>), footing location (<i>L/B</i>), slope angle (<i>β</i>), slope height (<i>H/B</i>), and anisotropic strength ratio (<i>r</i><sub><i>e</i></sub>). Sensitivity analysis showed that <i>r</i><sub><i>e</i></sub>, <i>s</i><sub><i>uc</i></sub><i>/γB</i>, and <i>L/B</i> had a positive influence on capacity, while <i>β</i> had a negative effect, and <i>H/B</i> showed minimal impact. To enhance the predictive performance, a hybrid artificial intelligence model combining extreme deep factorization machine (xDeepFM) and beluga whale optimization (BWO) was developed. The xDeepFM model captured complex feature interactions, while BWO was applied to optimize the model’s hyperparameters. The hybrid model was trained and validated using the reference FELA dataset, achieving high predictive accuracy with values of <i>R</i><sup>2</sup> of 0.984, RMSE of 0.084, and MAE of 0.068. This integrated approach offers an efficient and interpretable alternative to conventional numerical methods for evaluating undrained bearing capacity in anisotropic soils near slopes.</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

Bearing Capacity Prediction of a Strip Footing on Slope Using a Hybrid xDeepFM and Beluga Whale Optimization Algorithm

  • 期刊论文

  • 发表日期：

    2025年9月24日

查看全文
  • Nitchapat Pinnatsakda,
  • Nattakan Malai,
  • Katavut Vichai,
  • Wittaya Jitchaijaroen,
  • Mohammad Khajehzadeh,
  • Suraparb Keawsawasvong

摘要

Accurate estimation of undrained bearing capacity near slopes was crucial for geotechnical stability, particularly in anisotropic clay where the strength varied by direction. This study employed finite element limit analysis (FELA), based on the AUS failure criterion under plane strain conditions, to generate reference bearing capacity values. Normalized input variables were analyzed: undrained shear strength ratio (suc/γB), footing location (L/B), slope angle (β), slope height (H/B), and anisotropic strength ratio (re). Sensitivity analysis showed that re, suc/γB, and L/B had a positive influence on capacity, while β had a negative effect, and H/B showed minimal impact. To enhance the predictive performance, a hybrid artificial intelligence model combining extreme deep factorization machine (xDeepFM) and beluga whale optimization (BWO) was developed. The xDeepFM model captured complex feature interactions, while BWO was applied to optimize the model’s hyperparameters. The hybrid model was trained and validated using the reference FELA dataset, achieving high predictive accuracy with values of R2 of 0.984, RMSE of 0.084, and MAE of 0.068. This integrated approach offers an efficient and interpretable alternative to conventional numerical methods for evaluating undrained bearing capacity in anisotropic soils near slopes.