Abstract: Back-filling mining method with whole tailings as material can not only reduce the mining cost, but also recycle solid wastes. At the same time, filling solid waste underground protects environment and maintains ecological balance. Since whole tailing’s particle size is small, whole tailings shall be mix with rod-mill tailings and Gobi sand for mine filling. It is necessary to research the optimal proportion of mixed filling aggregate. First, we tested the particle-size gradation and the nonuniform coefficient of rod-mill tailings, Gobi sand and whole tailings. Then we carried out 9 sets of strength tests of cemented fillings with different mixing ratios. On this base, we built the artificial neural network model for strength predictions and trained it with the experimental samples. Finally, we predicted the strength of mixed filling aggregate of orthogonal design by using the prediction model, and we revealed the relationship between filling body’s strength and characteristic value of mixed filling aggregate by using range analysis and regression analysis. The research results show that, with different mixed aggregate’s average sizes and nonuniform coefficients, the filling body has significant strength differences between early and late stage. Mixed aggregate with smaller average particle size has a higher strength in early stage; while the larger is more inclined to increase the filling body’s strength in late stage.

Key words: mixed filling aggregate, average particle size, filling body’s strength, particle size gradation optimization, neural network