Scholarlink

<p>The growth of the assembly construction industry has led to a gradual increase in the demand for precast concrete. However, it is difficult to obtain a mix ratio for precast concrete with optimal properties. In this study, the optimal proportioning of precast concrete was investigated based on machine learning for practical engineering applications. Three models, artificial neural network, random forest and eXtreme Gradient Boosting (XGBoost), were trained and used to predict the strength of precast concrete. The predictions were evaluated using mean square error and R-squared error. The results showed that the trained XGBoost model presented the highest prediction accuracy for the compressive strength of precast concrete. The importance and effect of features on precast concrete were analyzed by SHapley Additive exPlanations. It was found that cement contributes to the initial mechanical properties of precast concrete, while blast furnace slag enhances the long-term mechanical properties of precast concrete. The optimal incorporation of raw materials was also discovered. The incorporation of cement and blast furnace slag will favor the development of strength of precast concrete, while fly ash was to be incorporated in an amount not more than 75&#xa0;kg/m<sup>3</sup>.</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

Prediction and in-depth analysis of precast concrete strength by machine learning

  • 期刊论文

  • 发表日期：

    2025年3月3日

查看全文
  • Jiarui Gu,
  • Chao Wei,
  • Shanliang Ma,
  • Jie Wang,
  • Yang Shao,
  • Zengqi Zhang,
  • Xiaoming Liu,
  • Lilei Zhu,
  • Chun Han

摘要

The growth of the assembly construction industry has led to a gradual increase in the demand for precast concrete. However, it is difficult to obtain a mix ratio for precast concrete with optimal properties. In this study, the optimal proportioning of precast concrete was investigated based on machine learning for practical engineering applications. Three models, artificial neural network, random forest and eXtreme Gradient Boosting (XGBoost), were trained and used to predict the strength of precast concrete. The predictions were evaluated using mean square error and R-squared error. The results showed that the trained XGBoost model presented the highest prediction accuracy for the compressive strength of precast concrete. The importance and effect of features on precast concrete were analyzed by SHapley Additive exPlanations. It was found that cement contributes to the initial mechanical properties of precast concrete, while blast furnace slag enhances the long-term mechanical properties of precast concrete. The optimal incorporation of raw materials was also discovered. The incorporation of cement and blast furnace slag will favor the development of strength of precast concrete, while fly ash was to be incorporated in an amount not more than 75 kg/m3.