Unsaturated Wet Granular Materials Rheology: A Study Using X-ray Micro-tomography and Deep Learning Models

This study presents experiments and discrete element simulations of simple, normal stress-controlled shear flows in model unsaturated granular materials, consisting of frictional spheres bonded by a small amount of wetting liquid. The rheology is characterized by the internal friction coefficient μ∗_w and solid fraction φ_S, as functions of reduced pressure P∗ and a visco-inertial number J = I_v + 2I². X-ray micro-tomography is used to explore the microstructure, with advanced segmentation techniques combining Random Forest and U-Net models for accurate phase distinction and quantification. A new automated tool classifies liquid morphologies, demonstrating shear-induced redistribution and clustering. Coordination number estimation further details interactions between solid particles and liquid phases, providing a robust methodology for quantitative analysis of wet granular material microstructures.