Discrete particle modeling has been performed to simulate failure of bonded rock (in 2D) at various confining pressures. While shear bands develop at low confinement, localization at high stress levels occur through formation of low-angle compaction band like features. Significant grain crushing is associated with the compaction bands, as in laboratory experiments. Compaction band development takes place primarily in periods of axial stress relief, and is associated with low angle shear band formation. This kind of modeling appears to be a promising tool for understanding strain localization mechanisms in general and in particular to bridge between field and laboratory observations of compaction bands.
Compaction bands have been observed experimentally in high porosity sandstone by several laboratories (e.g. by Olsson and Holcomb  and by Wong et al. . Bands observed in laboratory specimens are typically generated under high confining stress in triaxial tests, where they develop approximately perpendicular to the largest principal stress, and are associated with significant grain crushing. The laboratory observations show that they can be formed under strain hardening  as well as in a strain softening / ideally plastically deforming mode , and that they may grow by thickening  or formation of multiple bands . Gettemy and Holcomb  demonstrate from a study of acoustic emission focal mechanisms that low angle shearing is a key feature in the growth of compaction bands observed in Castlegate sandstone. Evidence of compaction bands in nature has been reported by Mollema and Antonellini  and by Sternlof et al. . Obviously, the conditions under which these bands were formed are not fully known, but stresses required to form bands in situ appear to be lower than in the laboratory. They are identified as possible permeability barriers, with significant pore volume reduction and also associated grain crushing, although comminution is less apparent in the field than in the laboratory. Multiple bands are observed, with a tendency of being thicker in the middle than towards the edges, showing signs of lateral growth, observation of which is naturally hampered in laboratory specimens. The theoretical understanding of compaction band formation stems primarily from the use of bifurcation analysis based on elastoplastic continuum theory [6, 7]. A basic requirement for localization is compaction behavior associated with a yield surface having an end cap. These theories, being macroscopic in nature, do however not give any insight into the microscopic mechanisms of band formation and growth. As pointed out in the summarizing article by Holcomb et al. , the compaction band phenomenon is not yet fully understood. In particular, grain scale mechanisms need to be addressed in order to understand under what conditions and in which rocks compaction bands can be formed, and further to be able to predict their evolution and consequences for e.g. petroleum reservoir performance. Discrete particle modeling is a powerful tool to study deformation and failure mechanisms of bonded granular media [9, 10] Using such methods, complex dynamic material behavior emerges from simple contact mechanics through cooperative processes in a manyparticle assembly.