Geological Influences on the Strength of Rocks and Implications on Brittle Rock Mass Behavior

Selection of the Hoek-Brown material constant, "mi" for use in design is often not based on thorough laboratory test results but rather on an estimate solely based on rock type from published tabulations. Rock strength properties are well known to have a high coefficient of variation and therefore a more rigorous approach to determining "mi" needs to be considered when laboratory testing data is not available, particularity for feasibility studies and preliminary design. This research examines a suggested approach that considers the interlocking nature of grains, modal grain size, anisotropy, porosity, and kurtosis or the degree of mixing of the grain to make estimates of the material constant. The procedure is applied to a number of rock sample photos, which were taken as part of a data set of compressive and tensile strength testing on a variety of rock types. The individual influences are examined and the resulting effects on the value of "mi" are discussed. In addition, the way in which such an approach can be useful in conventional numerical modelling techniques are explored. This approach, based on visual inspection of rock type characteristics, was found to have an overall accuracy of up to 81% when compared to mi values calculated from laboratory test data. The approach works best with granitic rocks showing an accuracy of 96%, however, it is less effective with finer grained rocks such as limestone with an accuracy of 59%. When visual "mi" parameters obtained by using the photo-analysis procedure on a limited dataset of sample pictures was applied in numerical modelling of a deep-seated tunnel. The visually determined mi resulted in a conservative overestimate of the depth of damage ranging from 4% up to 35% over a limited point load and Unconfined Compressive Strength test data set used to estimate the laboratory "mi" value. The approach can be used as an index assessment method that can be quickly executed in the lab, as well as in the field during sampling or core logging, potentially giving a more reliable estimate of the "mi" parameter than simply selecting a value from published tables based on rock type.