Scholarlink

The manufacturing of Concrete can be made more sustainable by using supplementary cementitious materials (SCMs), which provides an efficient way to partially substitute Cement. This study presents a comprehensive Neural Network based approach to predict the compressive strength of concrete for binary concrete mixes. The investigation incorporates various SCMs, such as fly ash, rice husk ash, silica fume, Alccofine, and Metakaolin, within a unified NN framework. Feedforward Backpropagation (FFBP) based Neural Network (NN) were employed to analyse an extensive dataset compiled from published research. To enhance predictive accuracy, binary-based one-hot encoding was applied, allowing integration of multiple SCMs into a single predictive model, eliminating the need for separate analyses. The FFBP-NN achives R2 values of 0.98869, underscoring its robustness in incorporating numerical and categorical variables. A comparative evaluation with existing models, including a biology-based evolutionary algorithm FFBP-NN, highlighted the critical role of higher training data and categorical variable representation in achieving accurate predictions. This study emphasizes the potential of categorical variable-based NN approaches to advance predictive modelling for concrete strength. Future research should focus on expanding input data and incorporating diverse materials and environmental conditions, considering these as categorical variables to further enhance model applicability.

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

Unified Neural Network Model for Predicting Compressive Strength of Concrete with Various Supplementary Cementitious Materials

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Prasenjit Sanyal

摘要

The manufacturing of Concrete can be made more sustainable by using supplementary cementitious materials (SCMs), which provides an efficient way to partially substitute Cement. This study presents a comprehensive Neural Network based approach to predict the compressive strength of concrete for binary concrete mixes. The investigation incorporates various SCMs, such as fly ash, rice husk ash, silica fume, Alccofine, and Metakaolin, within a unified NN framework. Feedforward Backpropagation (FFBP) based Neural Network (NN) were employed to analyse an extensive dataset compiled from published research. To enhance predictive accuracy, binary-based one-hot encoding was applied, allowing integration of multiple SCMs into a single predictive model, eliminating the need for separate analyses. The FFBP-NN achives R2 values of 0.98869, underscoring its robustness in incorporating numerical and categorical variables. A comparative evaluation with existing models, including a biology-based evolutionary algorithm FFBP-NN, highlighted the critical role of higher training data and categorical variable representation in achieving accurate predictions. This study emphasizes the potential of categorical variable-based NN approaches to advance predictive modelling for concrete strength. Future research should focus on expanding input data and incorporating diverse materials and environmental conditions, considering these as categorical variables to further enhance model applicability.