Scholarlink

<p>This study utilises the machine learning approach to predict the dimensionless bearing capacity (UBC)<sub>p</sub> of square footing placed over geocell reinforced stratified foundation bed containing construction demolition waste infilled in the geocell reinforcement compacted at different relative densities (R<sub>d</sub> = 30%, 50%, and 70%) overlying the clay subgrade of different strengths (q = 5&#xa0;kPa and 50&#xa0;kPa). Machine Learning models can improve the prediction accuracy of complex and non-linear relationships between soil properties and bearing capacity that are often oversimplified in traditional analytical or empirical methods and reduced the dependence on extensive field testing. A total dataset of 54 cases obtained from the laboratory experimental study were used for the machine learning analysis. The input variables including the average Young’s modulus of elasticity (E<sub>avg</sub>) of both soil layer, average friction angle (ф<sub>avg</sub>), average unit weight (γ<sub>avg</sub>), average cohesion (c<sub>avg</sub>) and incremental cohesion (c<sub>r</sub>) due to the geocell reinforcement were used to predict the bearing capacity. In this study, five different machine learning models including decision tree (DT), k-nearest neighbour (KNN), artificial neural network (ANN), random forest (RF), extreme gradient boosting (XGB) were used for the analysis. The predicted results of bearing capacity by different machine learning models were accessed using relative comparison of R-square, root mean square error (RMSE), and variance accounted for (VAF) values. The relative comparison of different machine learning models result show that R-square value of different model for training and testing dataset varied from 0.79 to 1 and 0.72 to 0.96. The decision tree (DT) model and extreme gradient boosting (XGB) model gives the best performance (R<sup>2</sup> &gt; 90%) among the other machine learning models. Furthermore, the relative impact of each input variable on estimating the bearing capacity is computed by conducting sensitivity analysis.</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 Bearing Capacity of Geocell Reinforced Layered Foundation Bed Using Machine Learning Models

  • 期刊论文

  • 发表日期：

    2026年2月2日

查看全文
  • Ravi Kumar Sharma,
  • Gaurav Juneja

摘要

This study utilises the machine learning approach to predict the dimensionless bearing capacity (UBC)p of square footing placed over geocell reinforced stratified foundation bed containing construction demolition waste infilled in the geocell reinforcement compacted at different relative densities (Rd = 30%, 50%, and 70%) overlying the clay subgrade of different strengths (q = 5 kPa and 50 kPa). Machine Learning models can improve the prediction accuracy of complex and non-linear relationships between soil properties and bearing capacity that are often oversimplified in traditional analytical or empirical methods and reduced the dependence on extensive field testing. A total dataset of 54 cases obtained from the laboratory experimental study were used for the machine learning analysis. The input variables including the average Young’s modulus of elasticity (Eavg) of both soil layer, average friction angle (фavg), average unit weight (γavg), average cohesion (cavg) and incremental cohesion (cr) due to the geocell reinforcement were used to predict the bearing capacity. In this study, five different machine learning models including decision tree (DT), k-nearest neighbour (KNN), artificial neural network (ANN), random forest (RF), extreme gradient boosting (XGB) were used for the analysis. The predicted results of bearing capacity by different machine learning models were accessed using relative comparison of R-square, root mean square error (RMSE), and variance accounted for (VAF) values. The relative comparison of different machine learning models result show that R-square value of different model for training and testing dataset varied from 0.79 to 1 and 0.72 to 0.96. The decision tree (DT) model and extreme gradient boosting (XGB) model gives the best performance (R2 > 90%) among the other machine learning models. Furthermore, the relative impact of each input variable on estimating the bearing capacity is computed by conducting sensitivity analysis.