Predicting compressive strength of quarry waste-based geopolymer mortar using machine learning algorithms incorporating mix design and ultrasonic pulse velocity

The current study aimed to investigate the possibility of predicting the compressive strength of geopolymer mortar by mix design parameters, ultrasonic pulse velocity (UPV) and machine learning techniques. Here the geopolymer mortar is produced from eggshell ash and rice husk ash as precursors, NaOH solution as activator and quarry waste as fine aggregate. Twenty-seven combinations of geopolymer mix and a total of 189 mortar cubes were cast and tested for UPV and compressive strength. Seven different machine learning techniques were used to predict the compressive strength assessment tools: linear regression, artificial neural networks, boosted tree regression, random forest regression, K-Nearest Neighbor, support vector regression and XGboost. Among the diverse machine learning models evaluated in this study, XGboost exhibited remarkable efficacy in forecasting the compressive strength of geopolymer mortar. The investigation conducted using SHAP indicates that the concentration of UPV shows the most substantial influence on the prediction of compressive strength.