When tunnelling at great depth in hard brittle rock or when mining-induced stresses cause stress-fracturing of brittle rock, the resulting rock fragments cannot fill the original space. As during a rock blast, geometric bulking occurs when brittle rock is fractured, and the volume occupied by fractured rock is much larger. Near underground excavation, this volume increase causes convergences including floor heave because the fractured rock can only move into the excavation. Unfortunately, analytical tools such as the convergence confinement method (CCM) or the ground reaction curve (GRC) do not account for this rock mass bulking action. Similarly, numerical continuum model, while accounting for material dilation, do not account for the unidirectional expansion (bulking) of the fractured rock.

The purpose of this thesis is to combine semi-empirical relations of bulking, established based on field measurements and numerical discontinuum models, with the analytical GRC-method and with 2D numerical models (specifically Phase2™) to provide a means for estimating the impact of bulking on tunnel convergence. The outcome of this thesis therefore is to provide a means of bulking enhance convergence prediction by analytical and numerical solutions. This is presented for circular tunnels, to facilitate use of analytical solutions, in different rock mass types (plastic and brittle) and for various stress states (stress ratio k = 1 and 0.5) as well as for mining conditions with associated stress changes.

The examples presented in this thesis demonstrate that rock mass bulking in brittle rock often dominates tunnel convergence. It is also shown that bulking by extension failure primarily affects the shallow radial displacement profile (near the excavation wall) whereas shear-related bulking, if not suppressed by sufficient confinement, causes deeper-seated radial displacements.

The practical implication of this work is that rock support experiences significantly more radial strain and deformation than predicted by conventional analytical and numerical solutions. These models therefore tend to underestimate the straining of installed rock support.