Investigating the Slurry Fluidity and Strength Characteristics of Cemented Backfill and Strength Prediction Models by Developing Hybrid GA‑SVR and PSO‑SVR

The waste rock and tailings backfill into the mined-out areas are the most effective method for solving the environmental pollution and surface disasters for nonferrous metals mines. In practice, the success and availability of backfill operations are dependent on the slurry fluidity and the strength properties of cement backfill. The transport of the slurry through the pipeline to the designated backfilling area relies on its eximious flow properties, while the appropriate strength of the filling body ensures the safe operation of the stope. In this paper, the effects of cement and aggregate types on the slurry fluidity and strength characteristics of cemented backfill are studied in detail, which are often ignored in other pieces of literature. Diffusivity is used as an indicator to evaluate the slurry fluidity. Various slurries whose concentrations ranging from 70%, 73%, 75%, 78%, and 80% are made with different aggregate ratios and cement-sand ratios are tested. It has been shown that slurry fluidity is inversely related to its concentration, but 78% is the “stopping point” for the deterioration of fluidity. The addition of rod-milled sand improves or worsens the cemented backfill (CB) strength depending on the amount of rob-milled sand. The uniaxial compression experiment results on 216 CB specimens produced by different combinations of influencing variables showed that CB specimens made from cement with superior mechanical properties have a higher uniaxial compressive strength (σucs). It has been also found that the effect of aggregate ratio on the CB strength is not singular, but works in conjunction with the curing time and the cement-sand ratio. The longer the curing time and the higher the cement content, the higher the CB’s σucs. To avoid the time-consuming and costly problem of obtaining the strength of the CB from indoor experiments, an SVR model capable of predicting the uniaxial compression strength of CB specimens is proposed, which is optimized by genetic algorithm (GA) and particle swarm optimization (PSO) algorithm. The results of the three performance indexes (MAPE, MSE, and R2) show the superior performance of the GA-SVR and PSO-SVR models and the agreement of the predicted results with the experimental results, which indicate that these two models can accurately predict the σucs of CB.