ALBERT

Description: The appearance of clay in fractures is an important issue of applied geoscience as it not only affects the stability but also the flow paths through rocks. Forming a link between hydraulic, geochemical and mechanical processes, clay structures need to be thoroughly investigated. The growing importance of clay for waste disposal, petroleum research, geothermal exploration and geotechnical engineering necessitates tools to find and to characterize clay structures and clay minerals indirectly from geophysical measuring methods. Particularly, there is need for a technique enabling to map clay-rich zones from geophysical well logs acquired on-site in order to assess the mechanical and hydraulic properties of rocks. In this study, we present a neural network based method to map clay bearing fracture zones in crystalline facies. The study has been performed on the basis of geophysical and geological data acquired at the geothermal site of Soultz-sous-Forêts (France), in the granitic reservoir. A neural network was trained on geophysical logs from the fully cored exploration well EPS1. Calibration of the network was done on reference logs derived from the drill core. The effective calibration enabled the creation of synthetic clay content logs, which predict the clay amount in fractures along the well with 〉74 per cent accordance with a reference log. High clay contents could be located in faults, on which aseismic movements have been identified. The validation of this relationship destines the synthetic logs to help identifying potentially weak zones from geophysical logging methods. With application on non-cored wells, this tool can become a powerful means for assessing the probability of aseismic movements on faults caused by the presence of clay and estimating the hydraulic properties of fractures.