Prediction of OPC Cement Compressive Strength Based on Cement Chemical and Physical Parameters Using Machine Learning Techniques

Abstract

This study aims to develop machine learning models to predict the 28-day compressive strength of Ordinary Portland Cement (OPC) based on chemical and physical parameters. The ultra-competitive cement industry requires companies to innovate continuously, but the conventional testing process takes at least 28 days, making product customization inefficient. This research proposes using machine learning techniques to accelerate this process. The predictive parameters include chemical components (C3S, C2S, C4AF, SiO2, etc.) and physical properties (Blaine, Residue, LOI, etc.) of OPC cement. The modeling was performed using random forest, gradient boosting, and artificial neural network algorithms. Model performance was evaluated using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R-squared values. The study used 1,570 valid data points from cement quality testing at PT Semen Gresik. Results show that the random forest method provides the highest coefficient of determination of 0.856 with RMSE of 13.086 kg/cm² and MAE of 10.784 kg/cm². The most significant attributes affecting prediction are CaO, Insol, SiO2, MgO, Al2O3, and SO3. Performance can be further enhanced through hyperparameter tuning using grid search method, achieving a coefficient of determination of 0.976 with RMSE of 6.118 kg/cm² and MAE of 5.198 kg/cm². This research contributes to accelerating cement quality control processes and supports faster product development in the cement industry.