Variance of intact rock strength determined by triaxial tests

Abstract

Given the results of triaxial tests, regressions are commonly used to evaluate the parameters of failure criteria that model intact rock strength, and to perform statistical inferences used to evaluate characteristic values defined in EC7. These inferences are only possible if the basic hypothesis underlying regression are followed, namely homoscedasticity (constant variance of the errors). This paper presents analyses of 23 sets of triaxial tests, starting by the evaluation of the regression parameters of the Mohr-Coulomb linear criterion and of the Hoek-Brown non-linear criterion, followed by carrying several statistical tests to check the homoscedasticity null hypothesis of independence between the variance of intact rock strength and the confining stress.