Scholarlink

<p>This study investigates the application of machine learning (ML) for predicting the compressive strength of cement mortar, which is influenced by a combination of mix design parameters, sand characteristics, and curing time. A comprehensive dataset was utilised, incorporating factors such as the aggregate-to-binder ratio, water-to-binder ratio, detailed sand properties (including fine content, sand content, mean particle size, uniformity coefficient, and others), and curing time. The study applied several ML algorithms, including linear regression, artificial neural networks (ANN), K-nearest neighbors (KNN), random forest regression (RFR), support vector regression (SVR), and extreme gradient boosting (XGB). The models were trained using a 70:30 train-test split and evaluated through 10-fold cross-validation, with performance metrics including R², root mean square error (RMSE), and mean absolute error (MAE). The results revealed that XGB performed the best, achieving a testing R² of 0.94 and an RMSE of 3.75&#xa0;MPa. This was followed by ANN with a testing R² of 0.93 and RMSE of 3.96&#xa0;MPa, and SVR with an R² of 0.92 and RMSE of 4.04&#xa0;MPa. Sensitivity analysis indicated that the aggregate-to-binder ratio (Agg/B), water-to-binder ratio (W/B), and curing time were the most critical factors affecting compressive strength. Although less influential, sand properties still contributed significantly, particularly in terms of aggregate packing and paste demand. The findings suggest that ML offers a promising approach to optimising cement mortar mix designs, enhancing quality control, and reducing the need for extensive experimental trials in the construction industry.</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

Machine learning-based prediction of cement mortar compressive strength: influence of mix proportions and aggregate characteristics

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月5日

查看全文
  • Navaratnarajah Sathiparan

摘要

This study investigates the application of machine learning (ML) for predicting the compressive strength of cement mortar, which is influenced by a combination of mix design parameters, sand characteristics, and curing time. A comprehensive dataset was utilised, incorporating factors such as the aggregate-to-binder ratio, water-to-binder ratio, detailed sand properties (including fine content, sand content, mean particle size, uniformity coefficient, and others), and curing time. The study applied several ML algorithms, including linear regression, artificial neural networks (ANN), K-nearest neighbors (KNN), random forest regression (RFR), support vector regression (SVR), and extreme gradient boosting (XGB). The models were trained using a 70:30 train-test split and evaluated through 10-fold cross-validation, with performance metrics including R², root mean square error (RMSE), and mean absolute error (MAE). The results revealed that XGB performed the best, achieving a testing R² of 0.94 and an RMSE of 3.75 MPa. This was followed by ANN with a testing R² of 0.93 and RMSE of 3.96 MPa, and SVR with an R² of 0.92 and RMSE of 4.04 MPa. Sensitivity analysis indicated that the aggregate-to-binder ratio (Agg/B), water-to-binder ratio (W/B), and curing time were the most critical factors affecting compressive strength. Although less influential, sand properties still contributed significantly, particularly in terms of aggregate packing and paste demand. The findings suggest that ML offers a promising approach to optimising cement mortar mix designs, enhancing quality control, and reducing the need for extensive experimental trials in the construction industry.