Scholarlink

<p>This study proposes a state-of-the-art deep neural network (DNN) and genetic programming (GP)-based methodology for the prediction of the CS of metakaolin mortar. The growing use of metakaolin mortar as a cement substitute reduces emissions and enhances concrete properties, but the complexity of blends hinders reliable CS prediction. Traditional methods are costly and slow, whereas modern soft-computing models offer efficient, accurate solutions. A comprehensive 421 experimental database was compiled for the literature, incorporating variables such as the cement grade, age, water-to-binder ratio, sand particle size distribution, and plasticizer amount. The performance of the proposed models was assessed using several performance indices. Additionally, the associated errors of the proposed models are plotted as an error histogram. The findings underscore the potential of DNN and GP models in metakaolin mortar design and optimization, highlighting their utility in understanding the influence of mix parameters on CS. The DNN is concluded to be the best-simulated model in the study; however, the performance of GP is also encouraging and provides an edge with its user-friendly empirical expression.</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

Deep Learning and Genetic Programming-Based Soft-Computing Prediction Models for Metakaolin Mortar

  • 期刊论文

  • 发表日期：

    2025年1月14日

查看全文
  • Manish Kumar,
  • Divesh Ranjan Kumar,
  • Warit Wipulanusat,
  • Sarawut Ramjan,
  • Akash Sankar Chowdhury,
  • Shreya Mazumadar

摘要

This study proposes a state-of-the-art deep neural network (DNN) and genetic programming (GP)-based methodology for the prediction of the CS of metakaolin mortar. The growing use of metakaolin mortar as a cement substitute reduces emissions and enhances concrete properties, but the complexity of blends hinders reliable CS prediction. Traditional methods are costly and slow, whereas modern soft-computing models offer efficient, accurate solutions. A comprehensive 421 experimental database was compiled for the literature, incorporating variables such as the cement grade, age, water-to-binder ratio, sand particle size distribution, and plasticizer amount. The performance of the proposed models was assessed using several performance indices. Additionally, the associated errors of the proposed models are plotted as an error histogram. The findings underscore the potential of DNN and GP models in metakaolin mortar design and optimization, highlighting their utility in understanding the influence of mix parameters on CS. The DNN is concluded to be the best-simulated model in the study; however, the performance of GP is also encouraging and provides an edge with its user-friendly empirical expression.