Tunnels excavated in massive unfractured Lac du Bonnet granite at the 420 Level of the Underground Research Laboratory showed typical signs of instability, i.e., spalling, slabbing, notch development. Two-dimensional elastic stress analyses of the failed tunnels indicated that failure was occurring at stress levels of about 100 MPa. A laboratory testing program was carried out using conventional unconfined-compression tests and triaxial-compression tests. In addition, uniaxial and biaxial tests of physical models containing circular holes were also conducted. All the laboratory tests indicated that the laboratory strength of Lac du Bonnet granite was twice the calculated stress level at which failure occurred. This result led to a study of massive rock strength using laboratory data and field observations.

In classical geotechnics, the shear strength of a rock is regarded as made up of two components, intrinsic strength or cohesion, and frictional strength. Laboratory tests were carried out which showed that initially, when rock deformations are essentially elastic, there is a maximum cohesive strength which is about 0.7-0.8 of the standard laboratory unconfined compressive strength. As the loads increase above this maximum cohesive strength, friction is increasingly mobilized and the associated nonelastic displacements damage the cohesion. Consequently at displacements near the peak strength, i.e., when friction is fully mobilized, approximately 70% of the initial cohesion has been lost. The laboratory tests showed that most of the cohesion loss results from very small displacements. The loss in cohesion was modelled using the Griffith locus based on a sliding crack model.

Microseismic monitoring of a circular test tunnel excavated on the 420 Level of the Underground Research Laboratory revealed that considerable damage, i.e., cracking, was iii occurring near the face of the advancing tunnel. Three dimensional numerical stress analyses were carried out to investigate the loading path near the face of the test tunnel. The analyses showed that the loading path exceeded the crack initiation stress measured in the laboratory tests but that the loading path did not exceed the initial cohesion values measured in the laboratory tests. Thus the loading path stress magnitudes were not sufficient to mobilize friction.

The effect of stress rotation, near the advancing tunnel face, was also investigated since it has been demonstrated, for tensile loading, that cracks can be made to grow at a constant load by rotating the direction of the applied load. Three-dimensional numerical stress analyses showed that the principal stress directions near the face of the tunnel rotate as the tunnel advances. It is proposed that the rotation of the stresses near the face amplifies the damage in situ and that this damage is equivalent to the damage in the laboratory tests in which the cohesion was reduced after only small displacements by 70%. Thus, when the maximum principal stress magnitude is above the crack initiation stress, the maximum cohesion in situ, that can be relied on, is only 50% or more of that measured in the laboratory.

Two-dimensional modelling of the failure process was carried out using a phenomeno- logical approach and a discrete fracture approach. The phenomenological approach uses a degraded strength to simulate the damage that has occurred near the advancing tunnel face. This approach, although practical, has several limitations, the most significant of which is the assumption that the entire rock mass strength around the tunnel has been degraded. The discrete fracture approach was carried out using a preliminary version of the finite element code InSight2D designed to model fracture growth in compression. A major advantage of the discrete fracture method is that it does not require the rock mass strength to be degraded. This approach holds much promise and captures one of the key physical processes, i.e. slabbing, observed around failing underground openings in brittle rocks.

The major contributions to our understanding of strength and failure that have resulted iv from the investigations and analyses carried out during the course of this thesis are noted below:​

• Samples obtained from a pre-stressed medium were damaged by the sampling process when the far field stress exceeded about 0.1σ_e.
• The effects of scale, loading-rate and moisture on intact rock strength is minimal and do not explain the observed strength reduction investigated in this thesis.
• Maximum friction and maximum cohesion are not mobilized at the same displacements. By the time friction is fully mobilized a significant portion of the maximum cohesion in the sample has been lost. The Griffith locus was successfully used to model the mobilization of friction and the loss of cohesion.
• Cracking around underground openings initiates at about the same stress level as crack initiation in the laboratory tests. The zone of crack initiation, with ϕ = 0, appears to define the limit of progressive failure around the openings at the 420 Level of the Underground Research Laboratory.
• The loading path of rock near the tunnel face suggests that stress rotation in the areas of maximum tangential stress concentration is a significant contributor to the localized degradation of the rock mass cohesive strength. Failure initiates in these locally damaged areas when the maximum tangential stress reaches about 100 MPa.
• Failure around the underground openings at the 420 Level of the Underground Re search Laboratory initiates at a point, called the process zone, and is progressive. Failure occurs as a slabbing mechanism and stops when the geometry of the notch becomes convex and the process zone becomes contained.
• Modelling of the extent of failure cannot be captured by plane-strain modelling unless some effort is made to account for the initiation of failure at a point near the tunnel face before plane strain conditions are reached.