Examining the Case for Accuracy and Precision When Determining the Hoek–Brown Parameters
摘要
The Hoek–Brown failure criterion has evolved from a pragmatic engineering tool into standard practice, yet its empirical foundations warrant renewed examination. This paper critically examines claims regarding "accurate" and "precise" determination of Hoek–Brown parameters (namely mi, mb, s, and GSI). In particular, while the criterion has proven useful over four decades, the foundational relationships for the mb and s parameters were derived from limited laboratory-scale testing of Panguna andesite and have never been validated through systematic field-scale rock mass testing. Our analysis reveals important considerations. First, identical mb values can arise from fundamentally different rock mass conditions, exposing a non-uniqueness problem. Second, GSI remains inherently qualitative despite attempts at quantification. Since mb and s are calculated from GSI, pursuing precise mi parameters cannot compensate for this fundamental uncertainty. Third, synthetic rock mass (SRM) modelling validates the Hoek–Brown framework while simultaneously indicating that mi varies with fracture intensity, suggesting that for naturally jointed rock masses, mi may not remain constant as conventionally assumed. Finally, the discussion examines how human factors facilitate the transition from "credible guide" to “established practice” and “standards” and how the use of AI tools risks amplifying unvalidated assumptions. The path forward requires transparency about empirical uncertainties, renewed field-scale validation, and recognition that computational sophistication cannot substitute for the reliable rock mass scale data that Hoek identified as critical components of rock engineering design.
Highlights Hoek–Brown relationships lack proper field-scale validation despite four decades of application in rock engineering practice. Identical mb values can represent fundamentally different rock mass conditions, creating a non-uniqueness problem. Pursuing decimal-figures precision for mi is inconsistent with qualitative GSI estimation that dominates mb uncertainty. GSI quantification attempts do not truly reduce uncertainty. AI tools risk amplifying unvalidated assumptions without addressing fundamental empirical limitations.