ABSTRACT

The prior knowledge of the rock mass behavior along a projected roadway is fundamental for planning activities and safety measures at a construction site. However, pre-investigations are often costly and time-consuming. To generate high resolution images of geotechnically important structures and changes in the rock mass, the Integrated Seismic Imaging System (ISIS) was developed at the GFZ. Seismic measurements offer detailed information on the rock mass, especially if the data acquisition takes place on-site during tunneling. However, to be of importance for the decision making on-site, the data needs to be processed and interpreted within a small timeframe. To meet this requirement the interpretation process needed to be automated. In the ONSITE project, a first step towards automating this process has been done by developing adapted routines with self-learning algorithms for rock mass classification based on seismic measurements. For the classification, the widely known RMR and RQD have been used so that a general idea about the rock mass behavior and not only single parameters can be gained from the results Based on the RMR, two rock mass classes were determined along seven seismic profiles from the Faido adit that belongs to the Gotthard base tunnel. The boundary between those classes was at 60 RMR which separates "fair" from "good" rock in the classification scheme. The RQD was separated into 3 classes, based on the number of occurrences, with either values in the range "excellent" (RQD>90), "good" (RQD 75 to 90) or "lower" (RQD<75). Both classification approaches using SVMs showed good training and testing accuracies, though the RQD was not as sensitive to the seismic velocities as had been expected.

This content is only available via PDF.
You can access this article if you purchase or spend a download.