Tetrahedral mesh generation can be used as a dynamic volume representation that can fully capture the complexity of a geological model and be adapted locally or globally to problems. It allows fine-grained control of mesh anisotropy and heterogeneity. An unstructured mesh generation robustness test was performed on input data containing 532 random fractures and 3 faults, with random angles and intersections and no preprocessing, resulting in a mesh of around 300k high-quality elements representing all fractures as triangular facets integrated into the final mesh.