Abstract: To solve the issue of solid waste resource utilization in a nickel mine in Northwest China,this study developed solid waste based wet shotcrete (SWC) for underground rock support by using waste rock and Gobi sand as mix aggregate and synergistically utilizing steel slag,blast furnace slag,desulfurization gypsum and other industrial solid wastes to prepare cementitious materials. The response surface method was used to design the experimental scheme,which systematically revealed the effect of different factors on the mechanical properties of SWC,and the formation mechanism of SWC with the help of various microscopic characterization methods. Meanwhile,an intelligent optimization model of SWC proportion was constructed by machine learning algorithm. The results show that the compressive strength of SWC increases and then decreases with the increase of dosage of steel slag (the compressive strength reaches a maximum of 35. 2 MPa when the dosage of steel slag is 30%),the ratio of rubber sand is positively correlated with the compressive strength,and the ratio of Gobi sand is negatively correlated with the compressive strength. The micro-mechanism study shows that the proportion of Gobi sand affects the pore structure of the aggregate stacking system (the aggregate stacking compactness is the largest when the proportion of Gobi sand is 0. 4),and the dosage of steel slag regulates the hydration reaction process of the cementitious system,and both of them affect the mechanical properties of SWC by changing the compactness of the material. The modeling results show that the parameter optimization of the support vector regression (SVR) model by Harris Hawks Optimization (HHO) significantly improves the prediction accuracy (R 2 >0. 99,RMSE<0. 20,VAF>99. 80),which is combined with Genetic Algorithm (GA). The HHO-SVR-GA intelligent model constructed by GA realized the accurate design of SWC ratio (error less than 5%). 

Key words: solid waste,underground support,wet shotcrete,machine learning,proportion design