Scholarlink

<p>Accurate prediction of compressive strength is vital for the design and application of high-performance concrete (HPC), where traditional empirical approaches often fail to capture the complex, non-linear interactions among mix constituents. In recent years, machine learning (ML) has emerged as a transformative tool, offering improved accuracy, adaptability, and efficiency in strength prediction. This review synthesizes recent advancements in the application of ML models, including Support Vector Machines, Decision Trees, Random Forests, Gradient Boosting, Extreme Gradient Boosting, and Artificial Neural Networks, for predicting HPC compressive strength. Key factors influencing prediction accuracy, such as mix proportions, curing conditions, environmental exposure, and the use of admixtures, are critically evaluated. Methodological considerations, including data preprocessing, feature selection, hyperparameter optimization, and model evaluation strategies, are discussed as essential for building robust predictive frameworks. In addition, emerging trends such as deep learning, hybrid meta-heuristic algorithms, IoT-enabled data collection, and explainable AI are highlighted for their potential to strengthen both reliability and transparency. Despite these advancements, research gaps remain, particularly the lack of external validation, dataset dependency, and the absence of standardized benchmarks. Overall, this review emphasizes the transformative role of ML in optimizing HPC applications and provides future directions for developing accurate, interpretable, and sustainable predictive models.</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

Advances in Machine Learning for Predicting Compressive Strength of High-Performance Concrete: A Review

  • 期刊论文

  • 发表日期：

    2026年5月28日

查看全文
  • Rupesh Kumar Tipu,
  • Sagar Paruthi

摘要

Accurate prediction of compressive strength is vital for the design and application of high-performance concrete (HPC), where traditional empirical approaches often fail to capture the complex, non-linear interactions among mix constituents. In recent years, machine learning (ML) has emerged as a transformative tool, offering improved accuracy, adaptability, and efficiency in strength prediction. This review synthesizes recent advancements in the application of ML models, including Support Vector Machines, Decision Trees, Random Forests, Gradient Boosting, Extreme Gradient Boosting, and Artificial Neural Networks, for predicting HPC compressive strength. Key factors influencing prediction accuracy, such as mix proportions, curing conditions, environmental exposure, and the use of admixtures, are critically evaluated. Methodological considerations, including data preprocessing, feature selection, hyperparameter optimization, and model evaluation strategies, are discussed as essential for building robust predictive frameworks. In addition, emerging trends such as deep learning, hybrid meta-heuristic algorithms, IoT-enabled data collection, and explainable AI are highlighted for their potential to strengthen both reliability and transparency. Despite these advancements, research gaps remain, particularly the lack of external validation, dataset dependency, and the absence of standardized benchmarks. Overall, this review emphasizes the transformative role of ML in optimizing HPC applications and provides future directions for developing accurate, interpretable, and sustainable predictive models.