Assessing Machine Learning Algorithms for Predicting Compressive Strength of Normal and High-Early Strength Concrete: A case Study in Binh Thuan, Viet Nam

Abstract

This study investigates the effectiveness of machine learning models in predicting the compressive strength of both normal and high-early strength concrete. The research, conducted as a case study in Binh Thuan, Vietnam, aims to address the challenges faced by engineers in optimizing concrete mix designs. The Support Vector Machine (SVM) model, in conjunction with regression analysis employing multilinear approaches, yields predictions that are comparatively less accurate than those obtained using deep learning methods such as Artificial Neural Networks (ANN) and Light Gradient Boosting Machine (LightGBM). Particularly, the ANN model exhibits superior predictive performance, boasting an impressive R-squared value of 0.988 and the lowest model error, measured by a Root Mean Square Error of 1.493. Moreover, these deep learning techniques prove adept at capturing the intricate relationship between the water-cementitious material ratio and concrete strength, thereby enhancing the effectiveness of quality control measures at the batching plant. Consequently, engineers are empowered to make precise adjustments to concrete mix proportions during the design phase, leading to a substantial improvement in prediction accuracy and ultimately ensuring the desired performance characteristics of the concrete.