Cross-Cuts GEOSTATISTICAL METHODS A GODSEND FOR ENGINEERS

When rock mechanic engineers try to design safe underground repositories for the storing of spent nuclear fuel bundles, they strive for sophisticated and reliable methods of predicting induced damage. Underground mine openings are usually riddled with faults, joints, slips and discontinuities. So using laboratory or in situ measurements of stress states around openings often leads to scattered results. When conventional statistical techniques are used to process these data, a mean value results, which is representative of the real rock property. But the approach taken by each engineer is often different, and conclusions vary.

Geostatistical methods, as opposed to conventional statistical methods, may offer the best solution yet. Developed in the late 1970s and early ’80s, geostatistics is based on the theory of regionalized variables and takes into account the correlation between two samples taken at a given distance apart.

Using data from an underground test performed by the Colorado School of Mines in an experimental mine near Idaho Springs, Colo., in 1981, J. Ouellet, Denis Gill and M. Soulie of Ecole Polytechnique in Montreal compared conventional statistical techniques with geostatistical techniques. The two results were very different.

The underground testing program consisted of 640 dilatometer tests in six rings of seven holes, each 6 m deep, drilled radially around an underground room. The rings were spaced at a distance of 2.4 m.

Based on the conventional statistical method of analysing the data, zones of induced damage seemed to extend to a maximum depth of 60 cm from the wall of the opening. Using the geostatistic approach, the Polytechnique team concluded that the dilatometer measurements do not detect any evidence of induced damage. “The induced damage zone, if it exists, is fewer than 20 cm in depth,” the researchers said. Their results were published in the August issue of Canadian Geotechnical Journal.