Scholarlink

<p>In this study the potential use of waste plastic as a sustainable partial substitute for fine aggregates (ranges from 0 to 20%) in M25-grade concrete were assessed. The study focused on both experimental outcomes and machine learning driven predictive modelling for predicting mechanical properties of sustainable concrete. To perform the study, experimental testing conducted on 189 specimens of each kind (cubes, cylinders, and beams) to evaluate their fresh, mechanical properties. Microstructural analyses employing scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDS) demonstrate matrix densification and compositional alterations at moderate substitution levels, with maximal enhancements noted at 10% partial replacement of WPD as fine aggregate. In contrast, flexural and splitting tensile strengths improved significantly up to an optimum replacement level of 10%, with increases of approximately 44% and 46%, respectively, compared to the control mix. SEM–EDS observations indicated improved particle packing, refined interfacial transition zones, and Ca–Si–rich hydration regions at moderate WPD contents, explaining the enhancement in tensile-related properties. Beyond 10% replacement, deterioration in workability and mechanical performance was observed. An XGBoost-GTO hybrid model was created to predict compressive, flexural, and splitting tensile strengths with good accuracy (R<sup>2</sup> = 0.9922, 0.9671, and 0.9639 in testing, respectively). Error indicators like RMSE, MAE, and WMAPE showed that the experimental and model predicted values were quite close to each other. The developed model achieved low AOC values for compressive strength (0.0125), flexural strength (0.0275), and splitting tensile strength (0.0228), confirming its robustness in accurately predicting the mechanical properties of sustainable concrete. The experimental and data-driven results show that using waste plastic in a controlled way at a 10% replacement rate not only makes structures stronger, but also, when combined with advanced machine learning modelling and easy-to-use prediction tools, creates a strong framework for smart and sustainable concrete design.</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

Metaheuristic-optimized XGBoost framework for predicting the mechanical strengths of WPP-modified high-performance concrete

  • 期刊论文

  • 发表日期：

    2026年2月19日

查看全文
  • Bhawesh Madhukar,
  • Sanjay Kumar,
  • Baboo Rai

摘要

In this study the potential use of waste plastic as a sustainable partial substitute for fine aggregates (ranges from 0 to 20%) in M25-grade concrete were assessed. The study focused on both experimental outcomes and machine learning driven predictive modelling for predicting mechanical properties of sustainable concrete. To perform the study, experimental testing conducted on 189 specimens of each kind (cubes, cylinders, and beams) to evaluate their fresh, mechanical properties. Microstructural analyses employing scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDS) demonstrate matrix densification and compositional alterations at moderate substitution levels, with maximal enhancements noted at 10% partial replacement of WPD as fine aggregate. In contrast, flexural and splitting tensile strengths improved significantly up to an optimum replacement level of 10%, with increases of approximately 44% and 46%, respectively, compared to the control mix. SEM–EDS observations indicated improved particle packing, refined interfacial transition zones, and Ca–Si–rich hydration regions at moderate WPD contents, explaining the enhancement in tensile-related properties. Beyond 10% replacement, deterioration in workability and mechanical performance was observed. An XGBoost-GTO hybrid model was created to predict compressive, flexural, and splitting tensile strengths with good accuracy (R2 = 0.9922, 0.9671, and 0.9639 in testing, respectively). Error indicators like RMSE, MAE, and WMAPE showed that the experimental and model predicted values were quite close to each other. The developed model achieved low AOC values for compressive strength (0.0125), flexural strength (0.0275), and splitting tensile strength (0.0228), confirming its robustness in accurately predicting the mechanical properties of sustainable concrete. The experimental and data-driven results show that using waste plastic in a controlled way at a 10% replacement rate not only makes structures stronger, but also, when combined with advanced machine learning modelling and easy-to-use prediction tools, creates a strong framework for smart and sustainable concrete design.