Abstract
In slope engineering it is quite normal to find that intact rock of a same unit shows high variability of strength and therefore variability in the rock mass properties. This is particularly true when dealing with rock that has been mineralized through hydrothermal processes. These deposits tend to produce marked alterations in the rock. Some types of alteration (e.g., silicified) could lead to a rock of higher strength (compared with the non-altered rock of the same unit), while other types (e.g., argillic alteration) result in lower strength. In general, veinlets, micro-defects and/or stockwork textures tend to decrease the uniaxial compressive strength of intact rock specimens, compared with the strength of fresh rock without such defects or texture. On the other hand, rock mass quality can also be affected by the alteration type, in terms of the degree of fracturing, joint condition and type of infilling material in the discontinuities. The type and degree of alteration affects the key geotechnical parameters used in rock mass classification systems and the slope stability analysis.
To address the problem of variability, regression analysis is commonly applied to the laboratory test results to establish a 'best fit' Hoek-Brown shear failure envelope. This fitting process, which in practice is normally accomplished using spreadsheets or commercial software, yields deterministic values of the Hoek-Brown parameters for the intact rock unit, without a quantification of the spread of measured values about the fitted shear strength envelope.
This paper revisits the problem of fitting a Hoek-Brown shear failure envelope to laboratory data, proposing a methodology that accounts result from unconfined compression strength tests to characterize intact rock units and direct shear tests to characterize joints shear strength.
In the proposed methodology, probability ranges for the computed Hoek-Brown parameters are determined, together with correlation coefficients that quantify the spread of values in the fitting process. Considering this approach, variability of the key parameters can be applied to calculate the probability of failure of a given slope design and slope stability analysis.
