A comparative study of machine learning techniques and data processing for predicting the compressive strength of pervious concrete with supplementary cementitious materials and chemical composition influence

Abstract

This study introduces a novel approach that combines machine learning algorithms, such as Extreme Gradient Boosting (XGB) and Artificial Neural Network (ANN), with chemical composition analysis to predict the compressive strength of pervious concrete. By considering a wider range of supplementary cementitious materials (SCMs) and chemical oxides, such as calcium oxide (CaO) and silicon dioxide (SiO₂), this approach significantly improves prediction accuracy over traditional empirical models, providing a more robust solution for sustainable construction. A comprehensive dataset of 659 observations was compiled from various studies, emphasizing the significance of input variables such as calcium oxide (CaO), silicon dioxide (SiO₂), aluminium oxide (Al₂O₃), and curing period. Various data processing methods were employed to enhance model performance, including Max-Min normalization, Z-score normalization, robust scaling, log transformation, and sigmoid normalization. The study demonstrates that XGB outperformed other machine learning models, achieving a training R² of 0.99 and a testing R² of 0.92, with an RMSE of 2.85 MPa. This research highlights the significance of incorporating chemical composition analysis (CaO, SiO₂) into machine learning models to enhance the prediction accuracy of the compressive strength of pervious concrete. The novelty of the approach lies in combining advanced data processing techniques with a diverse dataset of SCMs, offering an innovative solution for optimizing concrete formulations in engineering. Sensitivity analysis highlighted the critical importance of CaO, SiO₂, and curing period in predicting compressive strength, while aggregate size had a minimal impact. This research contributes to international efforts in sustainable infrastructure development by integrating machine learning techniques with chemical composition analysis to predict the compressive strength of pervious concrete. This innovative approach offers global implications for optimizing concrete mix designs, reducing material waste, and enhancing the durability of urban infrastructure.