Scholarlink

<p>This study presents a novel approach to improving the mechanical performance and durability of cement-stabilised soil by incorporating construction and demolition (C&amp;D) waste, polypropylene (PP) fibres, and sodium sulphate. While cement-stabilised soil is cost-effective for construction, its limited strength poses challenges for structural applications. To address this, 81 laboratory experiments were conducted to assess compressive and flexural strength across 5, 10, and 25-day curing periods. The inclusion of PP fibres significantly enhanced flexural strength, while C&amp;D waste and sodium sulphate markedly improved compressive strength. To predict strength outcomes, advanced machine learning models—a posterior broadcast neural network (PBNN) and random forest (RF)—were employed. Model parameters were optimised using the beetle antennae search algorithm with ten-fold cross-validation. The models demonstrated strong predictive performance, with R² values of 0.95–0.97 for PBNN and 0.93–0.99 for RF, and lower RMSEs compared to k-nearest neighbour, logistic regression, and multiple linear regression models Among the models tested, the Random Forest (RF) model demonstrated superior predictive accuracy, achieving a higher correlation coefficient (<i>R</i> = 0.99) and lower RMSE compared to the Backpropagation Neural Network (BPNN). This highlights the RF model’s effectiveness in capturing complex interactions in the dataset, making it more reliable for strength prediction of cement-stabilized soil composites. The integration of ML with experimental material design offers a data-driven strategy for developing sustainable, high-strength cement-stabilised soils.</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 strength characteristics of cement-stabilised soil incorporating construction and demolition waste and polypropylene fibres

  • 期刊论文

  • 发表日期：

    2025年11月1日

查看全文
  • Vivek Sivakumar,
  • Vedhasakthi K.,
  • Anurekha G.S.,
  • Kurra Hari Prasad,
  • Ravindaran Thangavel,
  • Poomalai R.

摘要

This study presents a novel approach to improving the mechanical performance and durability of cement-stabilised soil by incorporating construction and demolition (C&D) waste, polypropylene (PP) fibres, and sodium sulphate. While cement-stabilised soil is cost-effective for construction, its limited strength poses challenges for structural applications. To address this, 81 laboratory experiments were conducted to assess compressive and flexural strength across 5, 10, and 25-day curing periods. The inclusion of PP fibres significantly enhanced flexural strength, while C&D waste and sodium sulphate markedly improved compressive strength. To predict strength outcomes, advanced machine learning models—a posterior broadcast neural network (PBNN) and random forest (RF)—were employed. Model parameters were optimised using the beetle antennae search algorithm with ten-fold cross-validation. The models demonstrated strong predictive performance, with R² values of 0.95–0.97 for PBNN and 0.93–0.99 for RF, and lower RMSEs compared to k-nearest neighbour, logistic regression, and multiple linear regression models Among the models tested, the Random Forest (RF) model demonstrated superior predictive accuracy, achieving a higher correlation coefficient (R = 0.99) and lower RMSE compared to the Backpropagation Neural Network (BPNN). This highlights the RF model’s effectiveness in capturing complex interactions in the dataset, making it more reliable for strength prediction of cement-stabilized soil composites. The integration of ML with experimental material design offers a data-driven strategy for developing sustainable, high-strength cement-stabilised soils.