Adaptive drive-based integration technique for predicting rheological and mechanical properties of fresh gangue backfill slurry

The gangue grouting and backfilling (GGB) in the subsequent space of coal mining provides an effective way for green disposal of coal gangue. This study proposes a newly integrated intelligent model of mixing entropy- and congestion degree-based particle swarm optimization support vector regression (MC-PSO-SVR) to predict the yield stress, plastic viscosity, fluidity and uniaxial compressive strength (UCS) of fresh gangue backfill slurry (FGBS). Analysis demonstrates that the particle swarm optimal (PSO) algorithm based on adaptive adjustment strategy can effectively optimize the hyperparameters of support vector regression (SVR), and the MC-PSO-SVR model exhibits better predictive capability (R 2 > 0.88) and lower error coefficients (MAE, RSE, and RMSE values approaching 0) and narrower widths of 95 % confidence intervals for yield stress, plastic viscosity, fluidity, and UCS. Furthermore, the R 2 value surpasses 0.95 for external datasets, indicating enhanced generalization capability and robustness. Model significance analysis indicates that water content, 0–0.075 mm gangue sand (GS), and 0.075–0.15 mm GS are key factors controlling the rheological properties and mechanical strength of FGBS. The content of 0–0.075 mm GS and 0.075–0.15 mm GS has a significant positive effect on yield stress, plastic viscosity, and UCS, while showing a negative effect on fluidity. Nevertheless, the impact of water content exhibits a contrasting outcome. This is mainly because fine-grained (0–0.15 mm) GS improves the uniformity and density between particles in the slurry, while excess water content disrupts the balance provided by fine-grained GS, and the uneven distribution of particles in the slurry can be caused. This research facilitates the assessment and regulation of engineering properties of FGBS in practical engineering applications to meet different working conditions, providing a theoretical basis for further ratio optimization of backfill materials.