Influence of Sample Height-to-Width Ratios on Failure Mode for Rectangular Prism Samples of Hard Rock Loaded In Uniaxial Compression

Surface-parallel slabbing is a failure mode often observed in highly stressed hard rocks in underground excavations. This paper presents the results of experimental studies on slabbing failure of hard rock with different sample height-to-width ratios. The main purpose of this study was to find out the condition to create slabbing failure under uniaxial compression and to determine the slabbing strength of hard rock in the laboratory. Uniaxial compression tests were carried out using five groups of granite specimens. The mechanical parameters of the sample rock, Iddefjord granite from Norway, were measured on the cylindrical and Brazilian disc specimens. The transition of the failure mode was studied using rectangular prism specimens. The initiation and the propagation of slabbing fractures in specimens were identified by examining the relationship among the applied stress, strain and the acoustic emission. The stress thresholds identified were compared to those reported by other authors for crack initiation and brittle failure. It is observed that the macro failure mode will be transformed from shear to slabbing when the height/width ratio is reduced to 0.5 in the prism specimens under uniaxial compression. Micro σ ₁-parallel fractures initiate when the lateral strain departs from its linearity. Slabbing fractures are approximately parallel to the loading direction. Labotatory tests show that the slabbing strength (σ _sl) of hard rock is about 60% of its uniaxial compression strength. It means that if the maximum tangential stress surrounding an underground excavation reaches about the slabbing threshold, slabbing fractures may take place on the boundary of the excavation. Therefore, the best way to stop or eliminate slabbing failure is to control the excavation boundary to avoid the big stress concentration, so that the maximum tangential stress could be under the slabbing threshold.