A Novel Mesoscale Methodology for Predicting Frozen Wall Development Considering Groundwater Seepage

Artificial ground freezing (AGF) is widely applied in tunneling projects in highly permeable strata to form impermeable frozen walls and ensure construction safety. However, accurately predicting frozen wall development under groundwater seepage remains challenging due to the complex coupling between seepage and thermal fields and limited field data. This study proposes a novel mesoscale identification–reconstruction–Monte Carlo simulation methodology to predict the development of frozen walls under seepage conditions based on the actual geological features and seepage characteristics. The method involves the identification of sand microstructural characteristics, including porosity, particle ellipticity, and roughness, based on computed tomography imaging, realistic reconstruction of sand particles, and the establishment of coupled seepage–temperature finite-element models. Monte Carlo simulations are incorporated to account for the stochastic variability of sand characteristics, providing statistically significant predictions. The numerical models were validated through laboratory model tests simulating frozen wall development under different seepage velocities. Parametric analyses were further conducted to investigate the effects of freezing temperature, seepage pressure, porosity, particle ellipticity, and particle roughness on the freezing rate. The results reveal four key mechanisms governing frozen wall formation: the throat effect, the wall effect, the reflux effect, and the frozen soil limitation effect. This study enhances the mechanistic understanding of frozen wall development under seepage conditions and offers practical insights for optimizing AGF design and construction in permeable ground conditions.