Evaluation of artificial neural networks ability for complex aquifer modeling

Abstract:

Evaluation of groundwater level changes in Tabriz city area of eastern Azarbaijan is the aim of this paper. In many engineering projects, especially in Tabriz underground project, groundwater level changes are one of the main effective factors to these projects. Due to aquifer complexity in Tabriz area using of classical mathematical models has many problems. In this research different structures of artificial neural networks for water table forecasting of selected pizometer are used. Among the different structures, FFN-LM (Levenberg-Marquardt) is the best structure for this study. As well FNN-LM structure is used for forecasting of water levels in eight selected piezometers. This model Results were not useful, so the selected piezometers are divided into tow groups and prepared new models. In order to increase model efficiency, output nodes of model was increased equal to number of pizomemeters in any groups. These results could present monthly forecasting of selected piezometers water levels for two years interval.  

Keywords: Fluctuation of water table, artificial neural networks, Tabriz city area aquifer