Modified Stanley's approach for statistical analysis of compression strength test data of rock specimens

A very wide scatter is usually observed in laboratory compressive strength test (uniaxial and triaxial) data of rock specimens due to randomness in the number, orientation and distribution of micro-cracks. This leads to an uncertainty in choosing the representative design strength which leads to the need for a probabilistic approach to the analysis of test data. An attempt has been made in this paper to propose such an approach which is a modified version of Stanley's approach and employs Weibull's statistical strength theory. Uniaxial compressive strength (UCS) test data from three sources have been analyzed and results compared with those from Weibull's theory. As the proposed approach employs Weibull's parameters, goodness-of-fit tests have been performed to check the fitness of tests data to Weibull's distribution. Further, the proposed approach for uniaxial conditions has been extended to triaxial stress conditions. Corresponding cumulative distribution functions of the applied stress level have been obtained which have been subsequently invoked to correlate the applied stress level at failure with the associated risk of failure. For the UCS test data, the proposed approach yields higher design strengths than does Weibull's theory when Weibull's m parameter is greater than unity. Strength under triaxial stress condition can be predicted by employing the proposed approach even though triaxial tests cannot be conducted, provided the UCS test data are available.