Current Limit States (or Load and Resistance Factor) Design require determination of characteristic values of geotechnical parameters as a fundamental step in the design process. Unfortunately, there seems to be no consensus in the rock engineering community on the interpretation of the definition of characteristic values, with the result that there is often confusion about the method to be used to identify them. This is particularly true for the characteristic triaxial strength of intact rock. In this paper we will explore alternative possible approaches to determining the characteristic triaxial strength of intact rock. For the purpose of this analysis, two published data sets of triaxial strength are used. In our analysis we fit the Hoek-Brown criterion to these data, and compare alternative methods of estimating a criterion that represents characteristic strength. The methods we use are i) reducing the fitted (mean) criterion by some multiple of the standard error of the fit; and ii) using quantile regression. The results of these analyses are compared in the context of Limit State Design, and the appropriateness of their application discussed.
With the introduction of limit states design (LSD), also known as load and resistance factor design (LRFD), into geotechnical engineering, designers are required to adopt new definitions of some customary design concepts. The characteristic values of resistance (strength) and actions (loads) are examples of such new definitions.
Current geotechnical LSD codes (e.g. [1, 2]) require determination of characteristic values of geotechnical parameters as a fundamental step in the design process. Although these design codes and their associated support manuals usually give general guidance on selecting characteristic values, there seems to be no consensus in the rock engineering community on the interpretation of these guidelines. An unfortunate result is that there is often confusion about the method to be used to identify characteristic values. This is particularly true for the characteristic triaxial strength of intact rock, where additional complications arise due to the phenomenon of strength being as a function of confining pressure. This seems to be a problem specific to geotechnical LSD, since the required strength property of materials involved in structural codes (e.g. timber, steel or concrete) is the uniaxial strength which can be characterized with a single probabilistic distribution. In this paper we first briefly review the theoretical basis of LSD characteristic values, and then review in more detail the definition of characteristic values given in EN-1997 . We use two previously published triaxial strength data sets [3, 4] to explore and compare alternative approaches — namely non-linear (Bayesian) regression and quantile regression — for obtaining characteristic triaxial strength curves. We conclude with observations regarding the development of characteristic curves for use in LSD, and conservative criteria for use in customary design methods