Rock mass behavior is controlled by the mechanical characteristics of intact rock and fracture networks. For modelling purposes, the mechanical behaviour of natural fractures can be ‘smeared’ into a continuum, represented by ubiquitous joint models, or explicitly simulated. Smeared continuums define constitutive behaviour empirically based on intact rock and jointing conditions (e.g. Generalized Hoek-Brown criterion). Ubiquitous joints and explicit representation require that intact materials and fractures, or fracture networks, be parameterized and assigned constitutive models individually.
The complexity of constitutive behaviour applied to a numerical simulation can vary drastically. Linear- elastic models are simplistic, relying only on material stiffness and the bounding stress conditions as input. Their execution and interpretation are straightforward and, despite the obvious constitutive limitations, they provide sufficiently reasonable output in many applications. Non-linear modelling of materials with complex behaviour (including yielding and failure) requires the definition of peak and post-peak constitutive parameters. Post-peak parameterization substantially increases uncertainty in numerical models due to the limited reliability with which post-peak parameters can be deriving at the rock-mass scale.
Despite the exponential increase in uncertainty from linear to non-linear modelling, the trend in numerical simulations has been towards increasingly complex models. This trend is in part driven by the continuous progress towards easy-to-use, commercially available software packages. What practitioners routinely overlook, however, is that the degree of complexity incorporated in a numerical model must be justified by (1) the quality and quantity of geomechanical data available for material parameterization, (2) the degree of confidence in assumed boundary conditions, and (3) the rigor of calibration. Justification of complexity is required because numerical solutions to geomechanical analyses are often non-unique. In some cases, multiple combinations of input parameters and boundary conditions can result in the same output (final stresses and deformations). However, the path to achieving those final results may be substantially different and influence the fundamental interpretation of the underlying mechanisms. Further, when it is impossible to discern which combination of input values and boundary conditions are ‘correct’ for our back analysis of observations, the confidence placed on forward predictions is severely limited.
RockEng has extensive experience in all types of geomechanical numerical modelling with numerous software packages – this allows our team to ensure that the right tools are applied to any numerical modelling project. We believe in the practical and pragmatic application of numerical modelling to rock engineering problems and projects. To find out more about how our modelling experts can add value to your project – or to inquire about our services in numerical modelling courses and training – please contact [email protected]