Empirical approaches are now used in tunnelling works for decades. Several systems exist (Q system, RMR and others) that have parameters which describe the different components of the rock mass. They surely succeed in describing rock mass parameters including the matrix, fractures and, for some of them, external constraints. Thus, these systems allow describing various elements of rock masses, covering geological, mechanical and hydrogeological aspects to some extent. Because of the description of the rock forming elements and rock mass global rating, it is now common language to call them Rock Mass Classifications.
However, do Rock Mass Classifications truly classify the rock masses, as suggested by their name? For most rock mass classifications, the global scoring does not represent the rock mass by itself but the interactions of the rock mass on a considered underground opening (tunnel, cavern or mine). This direct link between a project and the characterization or scoring of a rock mass is neither obvious nor implicit in Rock Mass Classifications.
Therefore, many users consider that once the rock mass is rated, the rating can be used regardless of the project (span, orientation, depth, etc.). Such common misuse may explain why the Rock Mass Classification systems are widely criticised when their link to a specific project is where their real strength lies. Finally and even if too late, calling the empirical approaches "project rock mass classification" would have been more appropriate by highlighting the link between the rock mass and the specificities of the project on which they are applied.
Practical examples are provided to highlight these interactions and the subsequent contractual issues that may ensue, if rock masses are disconnected from engineering projects.
Apart from minor attempts, the very first system to estimate the quality of rock masses and classify them is the RQD (Deere 1964; Deere and Deere 1988). Most of today’s classifications already used this parameter, to estimate the intensity of jointing. The development of empirical approaches started during the 70’s with the RMR (Bieniawski 1973) and the Q-system (Barton et al. 1974). Followed by other systems - the Chinese BQ (in Feng and Hudson 2011), the GSI (Marinos and Hoek 2000), the RMQR (Aydan et al. 2014), etc. - more than 30 systems, classifications or adaptations of existing classifications can be found today. However, the two initial classifications, updated (Bieniawski 1989; Barton and Grimstad 1994) and now based upon much more cases, are still widely used, worldwide, along with the GSI.