Scholarlink

In this study, we applied machine learning techniques to classify concrete into consistency classes, using real-world data provided by an industrial partner in France. The dataset includes over 480 different mixtures and more than 1,515 concrete samples, most of which incorporate supplementary cementitious materials (SCMs) such as slag and fillers. There are twelve input features and one output representing each of the five consistency classes. The machine learning models demonstrated strong performance, with the Light Gradient Boosting Machine emerging as the best performer, with an F1-score exceeding 88%. Using Shapley Additive Explanations, we identified the water-to-binder ratio, admixture dosage, and binder content as the most influential factors affecting the slump. Overall, this research underscores the potential of machine learning to enhance concrete mix design and promote more sustainable construction practices, though future work is needed to validate the models across diverse regions, raw materials, and mix designs to ensure broader applicability and robustness.

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

Predictive Modeling of Consistency Class Using Machine Learning Techniques on Industrial Concrete Datasets

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Abdelhamid Hafidi,
  • Benoit Hilloulin,
  • Ilhame Harbouz,
  • Ahmed Loukili,
  • Ammar Yahia

摘要

In this study, we applied machine learning techniques to classify concrete into consistency classes, using real-world data provided by an industrial partner in France. The dataset includes over 480 different mixtures and more than 1,515 concrete samples, most of which incorporate supplementary cementitious materials (SCMs) such as slag and fillers. There are twelve input features and one output representing each of the five consistency classes. The machine learning models demonstrated strong performance, with the Light Gradient Boosting Machine emerging as the best performer, with an F1-score exceeding 88%. Using Shapley Additive Explanations, we identified the water-to-binder ratio, admixture dosage, and binder content as the most influential factors affecting the slump. Overall, this research underscores the potential of machine learning to enhance concrete mix design and promote more sustainable construction practices, though future work is needed to validate the models across diverse regions, raw materials, and mix designs to ensure broader applicability and robustness.