Uncertainty-aware modeling of grouting-induced pore clogging using a four-phase filtration framework and proxy-based simulation

Grouting in porous media is widely used for seepage control and ground improvement, but its performance is strongly affected by pore-structure variability, slurry properties, injection conditions, and deposition-induced clogging uncertainty. This study develops an uncertainty-aware framework by coupling a four-phase filtration–deposition model with an Adaptive Sparse Polynomial Chaos Expansion (ASPCE) surrogate. The deterministic model is validated using two independent laboratory infiltration tests, and uncertainty propagation is then evaluated under low, standard, and high uncertainty levels. A comparative analysis of linear and nonlinear dynamic deposition laws shows that the linear law predicts faster particle retention and earlier blocking, whereas the nonlinear law produces broader response distributions and stronger interaction-driven sensitivity. Probability density and cumulative distribution results indicate that the nonlinear law may generate non-unimodal terminal deposited-mass responses under high uncertainty. First-order and total-effect Sobol indices reveal that injection pressure dominates the early stage, while pore-structure, water–cement ratio, and deposition-related parameters become increasingly important during clogging evolution. Compared with brute-force Monte Carlo simulation, ASPCE greatly reduces computational cost while maintaining satisfactory predictive accuracy. The proposed framework provides an efficient basis for uncertainty-informed grouting parameter regulation and clogging risk assessment.