Alternative models for the calculation of the RMR and Q indexes for granite rock masses

Abstract

Empirical classification systems like the RMR and Q are often used in current practice of geotechnical structures design built in rock masses. They allow obtaining an overall description of the rock mass and the calculation, through analytical solutions, of strength and deformability parameters which are determinant in design. To be applied these systems need a set of geomechanical information that may not be available or can be difficult to obtain. In this work it is intended to develop new alternative regression models for the calculation of the RMR and Q indexes using less data than the original formulations and keeping a high accuracy level. It is also intended to have an insight of which parameters are the most important for the prediction of the indexes and in the rock masses behaviour. This study was carried out applying Data Mining techniques to a database of the empirical classification systems applications in a granite rock mass. Data Mining is a relatively new area of computer science which concerns with automatically find, simplify and summarize patterns and relationships within large databases. The used Data Mining techniques were the multiple regression and artificial neural networks. The developed models are able to predict the two geomechanical indexes using less information that in the original formulations with a good predictive capacity.