Permeability evolution and predictive modeling in grout-reinforced fractured rock masses: An intelligent computational approach

The permeability evolution characteristics of grout-reinforced fractured rock masses significantly influence the seepage stability control in underground engineering. In this study, fractured sandstone specimens under various confining pressures (0–10 MPa) were prepared using a high-pressure triaxial testing system. Reinforcement was performed using coal gangue-fly ash-based grout, and the permeability variations under effective stresses (1–9 MPa) were systematically measured before and after grouting. A permeability prediction model for grouted rock masses was established by employing swarm intelligence algorithms (Particle Swarm Optimization (PSO), Genetic Algorithm (GA), and Grey Wolf Optimizer (GWO)), and an interactive computational platform was developed. The results show that post-grouting permeability decreased by 54.83–99.75 % compared to pre-grouting values, exhibiting a power-law decline with increasing effective stress. Using six key factors—including fracture stress state, initial permeability, and grouting parameters—an 80-sample training dataset was constructed. A backpropagation (BP) neural network (6-7-1 topology) optimized by the GWO achieved high-precision permeability prediction (R² = 0.997, MAE = 0.051). Finally, a Python-based intelligent interactive computing system was developed, integrating parameter control, model computation, and result visualization. This provides theoretical support and technical tools for engineering grout design.