Scholarlink

<p>This systematic review explores the application of Artificial Intelligence (AI) in optimizing the mix design of fly ash-based geopolymer concrete (FABGC). Analyzing studies published between 2014 and 2025, it examines key methodologies, including machine learning models, optimization algorithms, and multi-criteria decision-making approaches. Critical aspects such as data preprocessing, AI model selection, hyperparameter tuning, explainable AI (XAI), and optimization strategies are synthesized to provide a comprehensive perspective on AI-driven FABGC research. The review identifies Deep Residual Networks (ResNet) and Extreme Gradient Boosting (XGB) as the most accurate models for predicting FABGC strength, consistently outperforming others due to their lower error metrics. Backpropagation Neural Networks (BPNN) and Adaptive Neuro-Fuzzy Inference Systems (ANFIS) also demonstrate competitive performance, while Random Forest (RF) and Decision Tree (DT) models excel in computational efficiency with shorter training times. Despite being the most widely implemented, Artificial Neural Networks (ANN) rarely achieve the highest predictive accuracy. Traditional regression methods, though straightforward, lag behind in performance. These findings underscore the need for standardized datasets, enhanced collaboration, and innovative AI-driven approaches to improve FABGC mix design optimization. Addressing these challenges will facilitate more reliable and efficient AI applications in sustainable concrete technology.</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

Artificial Intelligence in Geopolymer Concrete Mix Design: A Comprehensive Review of Techniques and Applications

  • 期刊论文

  • 发表日期：

    2025年5月13日

查看全文
  • Malik Mushthofa,
  • John Thedy,
  • Mochamad Teguh,
  • Purwanto,
  • Adjie Gemilang Pratama,
  • Ay Lie Han

摘要

This systematic review explores the application of Artificial Intelligence (AI) in optimizing the mix design of fly ash-based geopolymer concrete (FABGC). Analyzing studies published between 2014 and 2025, it examines key methodologies, including machine learning models, optimization algorithms, and multi-criteria decision-making approaches. Critical aspects such as data preprocessing, AI model selection, hyperparameter tuning, explainable AI (XAI), and optimization strategies are synthesized to provide a comprehensive perspective on AI-driven FABGC research. The review identifies Deep Residual Networks (ResNet) and Extreme Gradient Boosting (XGB) as the most accurate models for predicting FABGC strength, consistently outperforming others due to their lower error metrics. Backpropagation Neural Networks (BPNN) and Adaptive Neuro-Fuzzy Inference Systems (ANFIS) also demonstrate competitive performance, while Random Forest (RF) and Decision Tree (DT) models excel in computational efficiency with shorter training times. Despite being the most widely implemented, Artificial Neural Networks (ANN) rarely achieve the highest predictive accuracy. Traditional regression methods, though straightforward, lag behind in performance. These findings underscore the need for standardized datasets, enhanced collaboration, and innovative AI-driven approaches to improve FABGC mix design optimization. Addressing these challenges will facilitate more reliable and efficient AI applications in sustainable concrete technology.