Komodromos, MichailMichailKomodromosGorji, MahanMahanGorjiDüster, AlexanderAlexanderDüsterGrabe, JürgenJürgenGrabe2024-08-082024-08-082024-10-01Computers and Geotechnics 174: 106606 (2024)https://hdl.handle.net/11420/48691Mesoscopic approaches in material science aim to conceive multiphase materials as compositions of constituent phases, rather than homogenized media. X-ray Computed Tomography is a powerful tool that provides full field information of composites’ structure in the form of three-dimensional images. Image analysis combined with computational geometry concepts permit the transition from voxel images to adapted unstructured tetrahedral meshes, a conversion that delivers important computational benefits of image-based numerical simulations. The current article focuses on the case of cemented natural granular material of two cement saturation degrees. Actual cylindrical samples are scanned and reconstructed in the form of digital volumes. Image artefacts impede reliable particle segmentation. A novel image filter based on domain cylindrical decomposition and histogram compatibility enforcement is applied to alleviate the bias. The conversion of voxel images to quadratic tetrahedral meshes is proposed to facilitate accurate image-based numerical analysis. Localization of stresses is evident in the upgraded linear strain element configuration, but missing from that of constant strain element. The display of stress results and supplementary statistical analysis suggest that the granular skeleton sustains the load by forming intergranular stress channeling, while the cement matrix is partially activated at regions of granular contact to aid the stress flow.en0266-352X2024ElsevierCemented granular materialImage-adapted meshing techniqueImage analysisLinear strain elementsMesoscale FEMTechnology::620: EngineeringJournal Article10.1016/j.compgeo.2024.106606Journal Article