Determination of three dimensional hydraulic conductivities using a combined analytical/neural network model

Analysis of groundwater flow through fractured rock masses is an essential step in many engineering and environmental problems, such as in safety assessment of radioactive waste storages, hydrocarbon storage caverns and hydropower projects. The most important hydrological parameter in groundwater flow analysis is the hydraulic conductivity which is anisotropic and heterogeneous in the fractured rock masses. To analyze the groundwater flow correctly, some site investigations through boreholes must be carried out. One of the challenges in seepage analysis for an engineering project is how to determine the anisotropic and heterogeneous hydraulic conductivities of the fractured rock masses using the limited in situ investigation data. In this study, a new practical approach for the determination of three dimensional hydraulic conductivities of fractured rock masses is presented. Starting from rock fracture properties surveyed in six boreholes, the anisotropic hydraulic conductivities are estimated using the in situ injection test results and Oda's theoretical model. A neural network method is then utilized to generate the three dimensional heterogeneous hydraulic conductivities based on the anisotropic hydraulic conductivities along the six boreholes. In order to evaluate the reliability of this approach, a 3D numerical seepage model using code FLAC3D is performed for a real project. The inflow values in a shaft obtained with the 3D numerical analysis are compared with the in situ measured flow. The result indicates that the derived hydraulic conductivity is acceptable. (C) 2010 Elsevier Ltd. All rights reserved.