Scholarlink

<p>The utilization of marble powder as a partial cement replacement in concrete had gained increasing attention due to its environmental and economic benefits. However, an accurate prediction of the compressive strength of marble concrete remains a challenge due to the complex interactions between mix design parameters. The prediction of compressive strength is essential for optimizing marble concrete mix designs and supporting sustainable construction. This study employs a comprehensive experimental database to develop predictive models based on cement content, water-to-binder ratio, sand-to-gravel ratio, marble powder content, superplasticizer content, and curing age. Three modeling approaches were implemented: a Deep Neural Network optimized using a Genetic Algorithm (GA-ANN), a Decision Tree (DT) model, and an Artificial Neural Network evaluated under K-fold cross-validation (K-ANN) to enhance generalization performance. In addition, the Improved Grey Wolf Optimizer (I-GWO) was employed to refine mix proportions by integrating adaptive search mechanisms that enhance solution accuracy and convergence speed. The results obtained indicate that GA-ANN achieved superior accuracy, with R<sup>2</sup> = 0.99 compared to 0.97 and 0.89 for K-ANN and DT, respectively. Additionally, GA-ANN outperformed K-ANN and DT across all performance metrics, confirming its robust predictive capability. The results obtained also show that the I-GWO model identified an optimal mix with 40% of marble powder replacement, achieving the highest compressive strength of 94.95 MPa at 90 days.</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

Prediction and optimization of the compressive strength of marble powder-based concrete using AI techniques: machine learning and metaheuristic approaches

  • 期刊论文

  • 发表日期：

    2025年9月10日

查看全文
  • Yasmina Kellouche,
  • Rachid Djebien,
  • Aissa Laouissi,
  • Yacine Karmi,
  • Mostefa Hani,
  • Bassam A. Tayeh,
  • Yazid Chetbani

摘要

The utilization of marble powder as a partial cement replacement in concrete had gained increasing attention due to its environmental and economic benefits. However, an accurate prediction of the compressive strength of marble concrete remains a challenge due to the complex interactions between mix design parameters. The prediction of compressive strength is essential for optimizing marble concrete mix designs and supporting sustainable construction. This study employs a comprehensive experimental database to develop predictive models based on cement content, water-to-binder ratio, sand-to-gravel ratio, marble powder content, superplasticizer content, and curing age. Three modeling approaches were implemented: a Deep Neural Network optimized using a Genetic Algorithm (GA-ANN), a Decision Tree (DT) model, and an Artificial Neural Network evaluated under K-fold cross-validation (K-ANN) to enhance generalization performance. In addition, the Improved Grey Wolf Optimizer (I-GWO) was employed to refine mix proportions by integrating adaptive search mechanisms that enhance solution accuracy and convergence speed. The results obtained indicate that GA-ANN achieved superior accuracy, with R2 = 0.99 compared to 0.97 and 0.89 for K-ANN and DT, respectively. Additionally, GA-ANN outperformed K-ANN and DT across all performance metrics, confirming its robust predictive capability. The results obtained also show that the I-GWO model identified an optimal mix with 40% of marble powder replacement, achieving the highest compressive strength of 94.95 MPa at 90 days.