Quantitative analysis of sheared unsaturated wet granular materials using X-ray micro-tomography and advanced segmentation techniques

Abstract The microstructure of sheared unsaturated wet granular materials, comprising solid particles, liquid phases, and void spaces, is explored using X-ray micro-tomography. Advanced segmentation techniques are employed to overcome challenges in distinguishing phases within the material, utilizing a combination of Random Forest and U-Net models for accurate segmentation of the X-ray images. This methodology enables the quantification of the solid and liquid fractions within the sample, revealing the effects of shear deformation on their distribution. Additionally, an automated tool is designed to characterize the local geometry of small liquid domains, classified according to the number of connected liquid bridges joining grain pairs and the shape of such clusters. It is shown that deformation redistributes the liquid phase, which tends to be excluded from the strongly sheared regions. Coordination number estimates agree with published numerical simulation results. The study also addresses some limitations related to voxel size. The robust tools to analyse complex three-phase microstructure of wet granular materials are expected to improve the modeling of their rheology under different conditions. Graphical Abstract