Interplay of fabric anisotropy and drainage direction in soil–water retention behaviour of sand-silt mixtures

The soil–water retention curve (SWRC) is commonly treated as a material function in unsaturated soil analyses, although it may be affected by soil fabric and the drainage boundary condition imposed during laboratory testing. This study employs a coupled Discrete Element Method–Pore Network Model (DEM–PNM) framework to investigate the combined influence of fabric anisotropy and drainage direction on the water-retention behaviour of sand-silt mixtures. The model is first verified against experimental SWRCs reported for specimens subjected to different stress ratios. A systematic numerical program is then conducted on seven sand-silt mixtures with different grain-size distributions and three relative densities. Cylindrical specimens are prepared under different triaxial stress ratios to generate different stress-induced fabric states and pore-network anisotropy. Drying SWRCs are then simulated under three drainage configurations: one-dimensional bottom drainage, bottom-plus-lateral drainage, and full all-directional drainage. The results show that the pore-size distribution index, λ, remains largely insensitive to stress ratio and drainage condition. In contrast, the air-entry value (AEV) is affected by fabric anisotropy when drainage is restricted to a one-dimensional path, particularly when the major principal stress direction during specimen preparation is aligned with the drainage direction, a configuration commonly encountered in laboratory SWRC tests. For the lateral or full all-directional drainage configurations, numerical simulations indicate the effect of stress-induced anisotropy becomes negligible, as air invasion is governed by the most accessible boundary-connected pore pathways. Accordingly, a simple correction model is proposed to account for drainage-dependent AEV variation. A slope-stability example demonstrates that neglecting the interplay between fabric anisotropy and drainage-boundary effects may lead to overestimation of suction-induced shear strength and slope safety.