Perpustakaan Digital ITB

In most natural condition, hydraulic conductivity distribution is heterogeneous and anisotropic that is affected by local lithological condition, such as rock porosity and rock joint distribution. Therefore, the more porous of lithology the more hydraulic conductivity number it gets. In the previous study, spatial hydraulic conductivity distribution is modeled using Kriging with the aid of SeGMS software. Three dimensional (3D) hydraulic conductivity distributions in sedimentary rocks, which are isotropic and heterogeneous, can be used for groundwater flow modeling. This paper discusses the modeling 3D hydraulic conductivity distribution using Neural Network (NN). The hydraulic conductivity as a target value is trained segmentally from its position in x, y, z coordinate using NN. Numbers of nodes and hidden layers will be affected by complexity of the data. Geological validation and cross validation show that NN can be applied for modeling the spatial hydraulic conductivity distribution.