Some rock mechanics problems in correlating laboratory results with prototype reactions

Abstract

INTRODUCTION THE use of laboratory test results to predict the reaction of a rock massif encounters three major and inter-related obstacles. The first of these is the frequent reluctance of the user to recognize and accept the fact that rock is a heterogeneous material. It seldom is isotropic; it very often reacts as an anelastic medium ; it rarely occurs as a continuous medium. And, all too frequently the rock in situ may have natural stresses that cannot be quantitatively examined and thus cannot be simulated in the laboratory. The second major obstacle is that as yet we have no way to quantitatively assess the influence of gross geological defects such as faults, joints, fractures, folds, etc. Thus, we are unable to simulate their effects in laboratory testing. The third hurdle we face is the present lack of empirical data on the reaction of rock under a prototype load. We seldom design for failure as we try to introduce conservative safety factors when using laboratory data for design purposes. Thus, we seldom acquire quantitative information on prototype reactions. In other words, unless the prototype actually fails, its deformation may be too slight to be noticed unless proper instrumentation has been installed and a continuous record maintained of the prototype reaction. In my research on underground civil engineering structures, I found that such instrumentation was installed only on the following projects: Tumut-1 and Tumut-2 power stations in Australia, Poatina Power Station in Australia, Picote power station in Portugal, Clear Creek Tunnel in the United States, and a few scattered arch and gravity dams in the United States and Portugal. However, even if the data from these few projects disclosed good correlation between predictions based on laboratory test results and the actual rock reactions to the prototype load--and they do not f rom a statistical viewpoint it would appear that this is much too small a sample to be used for derivation of an accurate correlation procedure.