GPR abilities for the detection and characterisation of open fractures in a salt mine

Ground‐penetrating radar is widely used for mapping fractures in resistive rocks such as salt. In operating mines, fractures are often induced by the mining activity itself. Knowledge of the position and the characteristics of the fractures, and their evolution in space and time, are very important for mine safety. In the salt mine discussed in this paper, radar profiles are carried out daily to detect fractures in the roof of the galleries. For this study, radar measurements are carried out using two different systems at two different sites in the mine. Dynamic and static data are obtained at various antenna–roof distances. A supplementary profile is registered at 500 MHz. At the two sites, 15 boreholes are available and the fracture openings are measured. Most of the fractures observed in the boreholes can be observed on the radar profiles, except for the very thin fractures and the fractures of complex fracture systems, which cannot be resolved. The detectability of the fractures is analysed for the dynamic as well as the static data. Fractures larger than 10 mm can be detected at least up to a depth of 2 m (length of the boreholes). For fractures smaller than 10 mm, the detectability depends on the depth into the salt and the antenna–roof distance. These results may be site specific. In order to characterize the fracture opening, the frequency content of the radar reflection is analysed. Synthetic signals show that in order to obtain reliable results, the analysis has to be performed over a broad frequency range and not at one frequency. For the real data, a reference signal is required and some corrections have to be applied to the reflected wave to take into account the propagation, the geometry and the reflectivity of the roof. If the signal quality is good, the estimation of the openings by inversion is mostly satisfactory. For signals where this is not the case, the differences from the real openings can be explained by a spatial variation of the fractures, irregularities of the opening, the presence of fractures very close to each other or by the quality of the reference signal. It can also be due to the method itself, which is less efficient in some opening ranges. Although the estimation of the fracture opening is not perfect, the uncertainty about the opening is reduced. A statistical analysis should further improve the results.