A model study of rock-joint deformation

Abstract

Existing techniques of rock-joint modelling are reviewed. It is concluded that no methods presently in use are acceptable as either realistic models of mating rock joints or as mass production methods for the development of large, highly jointed models of rock masses. A method is described for producing mating tension fractures in a weak, brittle model material using a large guillotine device. Parallel sets of model joints can be produced which are continuous, cross-jointed or offset (stepped) depending upon the chronological order of fracturing. The direct shear properties of these three types are compared and evaluated. The model results are used as a basis for predicting the full-scale (1: 500) displacements accompanying shear failure of a 96-ft long prototype tension joint. Recent numerical modelling of jointed rock masses has been based on assumed values of the shear and normal stiffness of the joints. These components are found to dominate the elastic deformation properties of the intact rock. The results of shear and normal stiffness tests on the model joints are used for a careful assessment of these quantities. The shear stiffness (peak shear stress per unit tangential displacement) is found to be both normal stress and size dependent, and this is confirmed by a survey of shear-test data for joints in rock. There appears to be an inverse proportionality between test dimension and shear stiffness, for a given normal stress. The normal stiffness (normal stress per unit closure) is found to be dependent on the preconsolidation or virgin normal stress level. The problems of simulating the behaviour of jointed rock masses by the finite-element method are reviewed. Two particular drawbacks seem to be the conservation of energy demanded during computation, and the computer storage problems involved in modelling dilatent joints. Both these features are of fundamental importance to rock-mass deformation. A move towards realistic physical modelling is considered essential to an understanding of real processes.