Experimental studies and symbolic machine learning aided prediction model of the mechanical properties of recycled waste slurry micropowder mortar

Abstract

Ready-mixed concrete generates large quantities of concrete waste during production and transportation. The reuse of concrete recycled waste slurry in mortar and concrete is widely regarded as having significant potential and promising prospects. This study aims to investigate the impact of recycled waste slurry micropowder (RSP) on the microscopic structures and macroscopic mechanical properties of RSP mortar. A series of microscopic experiments and uniaxial compressive tests were conducted on RSP mortar at replacement rates of 0%, 10%, 20% and 30% of RSP for cement. The particle size, Scanning Electron Microscope (SEM), X-ray Diffraction (XRD) and Date Time Group (DTG) analyses of RSP show that RSP has weaker particle size uniformity, rougher surface, less mineral content and higher water content than cement, which leads to a lower comprehensive performance of RSP than cement. As the replacement ratio of RSP increases, the strength of RSP mortar gradually decreases. This study provides a new approach for the strength prediction of RSP mortar. The machine learning model based XGBoost and SHAP are introduced to quantify and rank the importance of the influencing factors of the compressive strength of RSP mortar. It is found that the influence of cement strength on the compressive strength of mortar is the most critical, followed by the importance of RSP replacement rate and particle size. Then, a symbolic regression is adopted to establish a prediction model for the compressive strength of RSP mortar, and the prediction process of pure data-driven method is transformed into a practical mathematical equation suitable for engineering application.