Scholarlink

<p>This study investigates the prediction of compressive strength in hollow concrete prisms (HCP) using three popular data-driven methods: Random Forest (RF), Adaptive Gradient Boosting (AGB), and Extreme Gradient Boosting (XGB). The models were trained using a dataset of 102 data points, incorporating four key input parameters: compressive strength of mortar (CSM), compressive strength of block (CSB), the ratio of height to thickness (h/t) of the hollow concrete prism, and the ratio of CSM to CSB. Based on analytical results and 1000 Monte Carlo simulations, the XGB model demonstrated impressive accuracy in predicting the compressive strength of HCP, with a coefficient of determination approaching 0.99 in both training and testing. XGB model required a minimum of 200 Monte Carlo simulation runs to ensure optimal predictive stability. Furthermore, Shapley Additive Explanation (SHAP) and two-dimensional Partial Dependence Plot (2D-PDP) analyses indicated that CSB is the most influential parameter, with its increase correlating positively with HCP compressive strength. An optimal design strategy to achieve compressive strengths between 25 and 30&#xa0;MPa was identified, suggesting CSM values between 15 and 25&#xa0;MPa, CSB between 30 and 40&#xa0;MPa, an h/t ratio between 2.00 and 3.00, and a CSM/CSB ratio between 0.10 and 0.50. To facilitate practical application, an online tool was developed, embedding the XGB model for accurate and accessible compressive strength predictions for engineers and students.</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

Investigation on compressive strength of hollow concrete prisms using data-driven approaches and parametric analyses

  • 期刊论文

  • 发表日期：

    2025年6月3日

查看全文
  • Duy-Liem Nguyen,
  • Tan-Duy Phan

摘要

This study investigates the prediction of compressive strength in hollow concrete prisms (HCP) using three popular data-driven methods: Random Forest (RF), Adaptive Gradient Boosting (AGB), and Extreme Gradient Boosting (XGB). The models were trained using a dataset of 102 data points, incorporating four key input parameters: compressive strength of mortar (CSM), compressive strength of block (CSB), the ratio of height to thickness (h/t) of the hollow concrete prism, and the ratio of CSM to CSB. Based on analytical results and 1000 Monte Carlo simulations, the XGB model demonstrated impressive accuracy in predicting the compressive strength of HCP, with a coefficient of determination approaching 0.99 in both training and testing. XGB model required a minimum of 200 Monte Carlo simulation runs to ensure optimal predictive stability. Furthermore, Shapley Additive Explanation (SHAP) and two-dimensional Partial Dependence Plot (2D-PDP) analyses indicated that CSB is the most influential parameter, with its increase correlating positively with HCP compressive strength. An optimal design strategy to achieve compressive strengths between 25 and 30 MPa was identified, suggesting CSM values between 15 and 25 MPa, CSB between 30 and 40 MPa, an h/t ratio between 2.00 and 3.00, and a CSM/CSB ratio between 0.10 and 0.50. To facilitate practical application, an online tool was developed, embedding the XGB model for accurate and accessible compressive strength predictions for engineers and students.