Speaker
Description
Evaporation-driven salt transport and precipitation in porous media is a complex multiphysics process affecting numerous natural and engineered systems, including salt-affected agricultural soils, porous building material degradation, saline aquifer CO2 storage, and solar-driven interfacial desalination. Dynamic pore-network models (DPNMs) can resolve these processes but suffer from severe time-step restrictions and high computational costs. Conversely, existing quasi-static pore-network models (QSPNMs), while computationally efficient, typically fail to capture solute convection driven by corner flow during drying and invasion events, and lack direct liquid-phase flux information necessary for accurate convective transport calculations. We develop a novel QSPNM framework that explicitly accounts for corner flow, solute transport, and salt precipitation along with their feedback effects. The model employs a time-splitting strategy where water vapor diffusion and solute diffusion are treated as time-dependent processes, while liquid redistribution and associated convective salt transport are represented as instantaneous capillary-driven redistribution events. A key innovation is our derivation of a time-integrated liquid flux approximation during these redistribution events using liquid mass conservation and post-redistribution throat conductances, enabling quantitative evaluation of convective solute transport.
The proposed QSPNM was rigorously validated against a fully implicit DPNM for both pure water and brine evaporation in one-, two-, and three-dimensional pore networks. Volume-weighted spatio-temporal absolute L2 errors remain below 0.02 for all quantities (liquid saturation, salt concentration, and precipitated salt) across all test cases, demonstrating excellent agreement. The time-integrated liquidflux approximation achieves median relative errors below 1% in 1D networks and below 10% in higher-dimensional networks when using post-invasion throat conductance. When using identical time steps (∆t = 0.01 s), the QSPNM is approximately one order of magnitude faster than the DPNM. Temporal convergence analysis demonstrates substantially improved numerical stability of the QSPNM compared to the DPNM. This robustness stems from treating liquid redistribution as instantaneous with physical invasion criteria rather than resolving transient dynamics, combined with fully implicit schemes for salt transport that ensure unconditional stability. We successfully applied the QSPNM to a large three-dimensional pore network (30 × 30 × 60 pore bodies, 54,000 pores) for both pure water and brine evaporation—scenarios that are computationally prohibitive for DPNMs. This demonstration confirms the framework’s capability to simulate realistic porous media approaching a representative elementary volume (REV), providing a pathway for developing robust upscaling strategies. The proposed QSPNM delivers substantial computational efficiency improvements over DPNMs by enabling much larger time steps while requiring significantly less computational time per step, without sacrificing accuracy. By preserving essential pore-scale physics while drastically reducing computational cost, the framework is well-suited for systematic parameter studies and uncertainty quantification on large pore networks, development of improved constitutive relationships for REV-scale continuum models and derivation of upscaling strategies for evaporation-driven salt precipitation.
| Country | Germany |
|---|---|
| Acceptance of the Terms & Conditions | Click here to agree |
