A Modeling Method for Predicting the Strength of Cemented Paste Backfill Based on a Combination of Aggregate Gradation Optimization and LSTM

Cemented paste backfill (CPB) is a sustainable mining technology that widely used in mines and helps to improve the mine environment. To investigate relationship between aggregate grading different affecting factors uniaxial compressive strength (UCS) of cemented (CPB), Talbol gradation theory neural networks evaluate determine optimum ratio. The mixed ratio with least amount cement (waste stone content river sand = 7:3) obtained by using pile compactness function combined experiments. In addition, response surface method design strength-specific UCS prediction model which uses LSTM considers aggregates have high accuracy. root mean square error (RMSE) results 0.0914, coefficient determination (R2 ) 0.9973 variance account for (VAF) 99.73. Compared back propagation network (BP-ANN), extreme learning machine (ELM) radial basis (RBF-ANN), can effectively characterize nonlinear individual predict sensitivity analysis on shows all 4 significant effect following ranking: (0.9264) > slurry concentration (0.9179) rock content) (0.9031) curing time (0.9031).