Structural and neural network analyses of fracture systems at the Äspö Hard Rock Laboratory, SE Sweden

Abstract: The > 10,000 fractures documented in the 450m deep Äspö Hard Rock Laboratory (HRL) provide a unique opportunity to study brittle deformation of a Swedish bedrock mass. The fracture population consists of six major sets, one subhorizontal and five sub-vertical. A classical structural analysis explored the interrelations between geometry and frequency of both dry and wet fractures with respect to depth and in-situ stresses. Three main findings are:In-situ stresses govern frequency distributions of dilated, hence water-bearingfractures. About 68.5% of subhorizontal fractures are dilated in the thrustregime above a depth of ca. 230m while 53% of sub-vertical fractures aredilated in the underlying wrench regime.Fractures curve both horizontally and vertically, a finding confirmed by theapplication of artificial neural networks that included Back-Propagation andSelf-Organizing (Kohonen) networks.The asymmetry of the total fracture population and tilts of the sub-Cambrianpeneplain demonstrates that multiple reactivations of fractures have tilted theÄspö rock mass 6° to the west.The potential space problem raised by this tilt is negated by systematic curvature of steep fractures, some of which sole out to gently dipping fracture zones. Fractures probably developed their curvature when they formed deep in crystalline crust in Precambrian times but have since reactivated at shallow depths. These findings add significantly to the conceptual model of Äspö and should be taken into account in future studies regarding the isolation of Sweden's high-grade radioactive waste in crystalline bedrock.