This paper firstly presents a review on applications of DEM in rock mechanics, with particular focus on similitude studies between DEM and physical experiments. A challenge in DEM simulations is to select appropriate micro-mechanical models (and parameters) to recover the observed macro-mechanical behavior. An ideal experiment would validate DEM simulations against identical physical models with similar micro-mechanical properties. The second part of the paper discusses the results of such investigations undertaken on numerical and physical cemented (cohesive) assemblies in uniaxial and triaxial compression. Physical samples were prepared using steel balls bonded with Portland cement, cured under controlled laboratory conditions and tested in uniaxial and triaxial compression. Uniaxial tests results were analyzed to characterize the damage thresholds with the help of acoustic emission and volumetric strain monitoring. Numerical simulations were conducted with PFC3D using known and derived micro mechanical parameters from physical testing. The results from both numerical and physical tests showed good correspondence in macroscopic behavior i.e. peak strength, stages of damage, mode of failures. However the numerical simulations reflected a stiffer mechanical response than physical assemblies.
The discrete nature of rock at both small and large scales results in complex mechanical behavior. At a micromechanical level rock is fundamentally anisotropic and inhomogeneous. At a macro-mechanical level rock has all the complexities of micro-mechanics as well as additional variations in mechanical properties and structure. As a result, the strength and mechanical behavior of rock depends strongly on many factors, not least stress path, discontinuity distribution and strength, and loading rate. Moreover the size and geometry of the sampled rock mass volume has been shown to influence strength and Deformation.
Numerous efforts have been made to study and predict rock mass behavior by analytical, empirical and numerical methods. For example, analytical techniques based on assumptions of linear elasticity and isotropy have been applied to simple geometric problems with mixed success [1]. By contrast, empirical techniques rely on past experiences, where the measured performance of a system is reconciled against measured parameters. Perhaps the most relevant example is the empirical derivation of the Hoek Brown rock mass strength criterion [2-5], which is widely accepted as a good starting point, but has limitations in its precision and accuracy.
Numerical methods continue to show promise, but there are many basic questions still left unanswered, in particular the strength and time dependent behavior of rock masses is far from fully understood.
Even so, with the advancement of computer technology, numerical methods have become increasingly popular. Over the last three decades the numerical methods applied in rock engineering are mainly Finite Element Methods (FEM), Boundary Element Methods (BEM) and Discrete Element Methods [1, 6]. In FEM and BEM, the continuum concept is applied, where the micromechanical properties of a system are assumed to have no effect on the macro-scopic behaviour. Although, BEM offers computational advantages over FEM for problems of elasticity, difficulties are encountered for problems with elasto-plasticity, non-linearity in stiffness, anisotropy, viscoplasticity, and micro to macro scale joints and fractures [7].
