ABSTRACT
The uniaxial, unconfined compressive strengths of Kayenta sandstone specimens, ranging over some six orders of magnitude and up to a cubic meter in volume, do not in a statistically significant way depend on sample size. The observed, but slight variations in the experimental results produced by laboratory and field tests are of the order of a few tens of percent and, most likely, reflect uncertainties in experimental procedures and environmental conditions, particularly moisture content. The published results of an equally extensive testing program on an altered igneous rock (quartz diorite) are compiled and reviewed to reveal an abrupt decrease in measured strengths between specimens tested in the laboratory and those tested in place at the field site. Furthermore, neither data subset shows a significant strength dependence on size. Differences in moisture content may account for some of the discrepancy in strengths, but no documentation exists to quantify this environmental effect. As observed in and confirmed by the field tests on sandstone, nonuniform loading may cause the strengths of quartz diorite field specimens to be underestimated.
INTRODUCTION
Size effect implies a certain degree of dependence of mechanical properties on volume; specifically, it refers to the volume dependence of unconfined, uniaxial strength of intact rock and other brittle materials, either in tension or compression. The variation in tensile strength of most materials is usually a decreasing function of volume; however, the degree of variation depends in large measure on experimental conditions, stress gradients, and geometric (shape) factors. For instance, a rather pronounced size effect on the tensile strength of rock is observed in practically all versions of the Brazilian test and in thick-walled hollow-cylinder tests. If, on the other hand, the applied tensile loads are as uniform as possible (e.g., the direct tensile test), the size effect becomes subdued and, in some instances, may not even be statistically significant (Wijk et al., 1978). With very few exceptions, the volume dependence of compressive strength of rock is slight. Brown (1971), in his review of the size-strength effects in rock, reiterates that there are three general types of size effects: namely, with increasing size compressive strengths either decrease, increase, or increase at first and then decrease. In equally general terms, the decreasing type of strength-size effect characterizes the behavior of hard, low-porosity igneous and some metamorphic rocks, whereas the soft, more porous sedimentary rocks may display an increasing strength-size effect. Observing all three types of size effect in as many major rock types, Obert et al. (1946) nevertheless conclude, on the basis of Tchebycheff's statistical theorem (ibid., pp. 56-59), that no size effect is apparent. To explain the decreasing size effect on tensile strength, one may seek out some version of statistical fracture mechanics. The simplest of these, due to Weibull (see Brown, 1971, or Ratigan, 1982), assumes a serial arrangement of elementary "volumes" and each possessing a strength within a certain distribution about the mean. The implied geometry is consistent with steel wires and glass filaments pulled in tension. To predict the results of similar tests on rocks, Brown (1971) and Pretorius (1972) propose that a more realistic approach would be to consider the strength distribution in parallel and series arrangements of elementary volumes in a composite.
