Characterizing soil micro-pore structure and hydraulic properties using DEM-PNM

Flow and transport in granular soils are governed by the microstructure of the pore space, which is determined by grain size distribution and their spatial arrangement. However, existing nondestructive characterization of soil pore structure cannot fully meet the multiscale representation, due to the high cost and low accuracy. In this study, a multiscale pore-structure reconstruction was proposed to combine the discrete element method (DEM) and the pore network model (PNM). Six soil samples were selected with different grain-size distributions. A systematic numerical procedure was also used for granular soils generation, multiscale pore-network construction, and flow property simulation. Granular soils were also applied with grain size distribution over multiple orders of magnitude. The soil grain-size distributions were divided into several scale intervals. While the ratio of the maximum to minimum particle size within each interval was controlled to be less than 10. Soil particles were generated with different gradations using DEM. The pore networks were extracted to construct a multiscale pore network model at different scales. Soil hydrodynamic properties were simulated after extraction. The results showed that the multi-interval DEM reconstruction accurately reproduced the particle distribution of soils with different gradations, when the grain size ratio within each interval was controlled below 10. Specifically, the coefficients of determination (R²) were not lower than 0.97 and root mean square errors (RMSE) were below 3%. The multiscale pore network covered a pore size range of 10³~10⁵ in magnitude. The pore-throat radius distributions generally followed a lognormal pattern. Compared with conventional single-scale reconstruction, there was a more continuous and complete representation of pore networks in granular materials with a wide particle-size range, thus compensating for the under-detection of the fine pores using X-ray computed tomography. The R² values between the predicted and measured water retention curves were higher than 0.97, and the RMSE values were lower than 1.37%, indicating that the multiscale pore network model accurately captured the structural effects of pore patterns on soil water retention. Multiscale DEM-PNM also reproduced differences in the transport properties among soils. Simulated intrinsic permeability ranged from 4.21 to 1852.68 μm², and relative gas diffusivity ranged from 0.11 to 0.32, indicating strong effects of pore structure and pore continuity on hydraulic transport. In the soils with available validation data, DEM-PNM predictions were closer to measured intrinsic permeability than CT-PNM ones. The low resolution of CT images was attributed to the fine pores, their underestimated pore connectivity and hydraulic transport. The multi-scale pore network was developed for the influence mechanisms of particle gradation on pore structure and macroscopic transport. Complex physical-chemical-biological reactions were systematically investigated at the pore scale.