Production and optimization of affordable artificial geopolymer aggregates containing crumb rubber, plastic waste, and granulated cork based on machine learning algorithms

This study investigated the effect of incorporating crumb rubber (CR), plastic waste (PW), and granulated cork (GC) as partial substitutes for fine aggregates on the density, compressive strength (CS), and water absorption of artificial geopolymer aggregates (AGAs). The properties of AGAs were evaluated, and the optimal utilization of waste materials was determined using response surface methodology (RSM) and machine learning models, including linear regression (LR), support vector regression (SVR), and Gaussian process regression (GPR). Additionally, the microstructure of the optimized samples was analyzed under varying curing temperatures (60 °C, 90 °C, and 500 °C). The results indicated that the inclusion of these waste materials led to a reduction in both the density and CS of the AGAs compared to the control mixtures. Among the machine learning models, the GPR model demonstrated superior performance, with R² values of 0.989 for density prediction and 0.937 for CS prediction. The GPR model consistently outperformed the LR, SVR, and RSM models, showing higher accuracy in predicting both density and CS. Microstructural analysis revealed significant differences in the AGAs cured at different temperatures, with denser microstructures observed at 60 °C and 90 °C compared to 500 °C. Exposure to higher temperatures resulted in the formation of microcracks and a decrease in the N-A-S-H gel content, leading to lower dry density values. The optimal blend of waste materials, comprising 146 kg/m³ of CR, 46.80 kg/m³ of PW, and 5.25 kg/m³ of GC, resulted in AGAs with minimum densities of 1268.17 kg/m³ , 1194.40 kg/m³ , and 732.80 kg/m³ at curing temperatures of 60 °C, 90 °C, and 500 °C, respectively. Additionally, these AGAs exhibited a CS of 4.95 MPa and a water absorption rate of 10.37 % after curing at 60 °C.