Strength prediction of silicon–aluminum phosphate geopolymers using machine learning

Silico-aluminate phosphate (SAP) geopolymers are three-dimensional amorphous polymers formed by the reaction of aluminosilicate with phosphate. They exhibit high strength, excellent thermal resistance, and corrosion resistance, making them highly promising for applications such as construction materials, fireproof/anticorrosive coatings, and hazardous waste immobilization. However, current research on SAP geopolymers is in its early stages, with limited reported studies and significant variability in optimal mix proportions for achieving high strength, which hinders further development. To clarify the factors influencing the strength of SAP geopolymers and advance their development, this study established, for the first time, a comprehensive database of seven key characteristic parameters and their corresponding strength values. Machine learning algorithms were used to developa model to predict the compressive strength of SAP geopolymers, along with a graphical user interface for strength prediction. The results showed that the random forest regression model delivered the best predictive performance (R2 > 0.8) and strong generalization capability. Increasing the curing temperature and phosphate concentration enhanced compressive strength. The molar ratios of phosphorus/aluminum, phosphorus/silicon, and silicon/aluminum molar ratio must be maintained within specific ranges to maximize strength. The optimal ranges for achieving peak strength were investigated.