The bonded-particle model (BPM) consisting of parallel-bonded disks or spheres suffers from the limitation that if one matches the unconfined-compressive strength ( ) of a typical hard rock, then the direct-tension strength ( ) of the model will be too large. This limitation can be overcome in two dimensions by introducing a polygonal grain structure to provide rotational restraint arising from inter-granular interlock. The flat-jointed BPM (in which each disk-disk contact simulates the behavior of a finite-length interface between two disks with locally flat notional surfaces such that even a fully broken interface continues to resist relative rotation) provides such a structure and supersedes the parallel-bonded BPM by mimicking more of the micro- and macro-mechanisms associated with rock damage.
The mechanical behavior of rock is controlled by its microstructure. Complex macroscopic behavior arises from microstructural interactions; thus, if one could replicate the microstructure and microstructural interactions within a model, then that model should reproduce the macroscopic behavior. Rock can be classified as either compact or porous. Compact rock has negligible porosity (e.g., igneous and low-porosity sedimentary — denoted here as hard rock) while porous rock has significant porosity (e.g., low-grade metamorphic and high-porosity sedimentary). The damage processes in rock either occur at the grain scale or their effects can be mapped to this scale; thus, a model microstructure of grains, interfaces and pore space should be sufficient to study failure behavior of rock in the brittle regime . This material has an inherent length scale that is related to disk size. If used directly, then it mimics compact rock; if used as a base material to which larger microstructural components are added (such as voids created by removing disks  or joints created by replacing contacts with smooth-joint contacts ), then it mimics porous or jointed rock.