An Innovative Approach to Determination of Double-Porosity Fractured Aquifers Hydraulic Parameters Using Artificial Neural Network

Abstract

Accurate determination of hydraulic parameter values is the first step to the sustainable development of an aquifer. Since Theis (1935), type curve matching technique (TCMT) has been used to estimate the aquifer parameters from pumping test data. The TCMT is subjected to graphical error. To eliminate the error an Artificial Neural Network (ANN) is developed as an alternative to the conventional TCMT by modeling the Bourdet-Gringaten’s well function for the determination of the fractured double porosity aquifer parameters. The neural network model is developed in a six-step protocol based on multi-layer perceptron (MLP) networks architecture and is trained for the well function of double porosity aquifers by the back propagation method and the Levenberg-Marquardt optimization algorithm. By applying the principal component analysis on the training input data and through a trial-and-error procedure the optimum structure of the network is fixed with the topology of [3×6×3]. The replicative, predictive and structural validity of the developed network are evaluated with synthetic and real field data. The developed network provides an automatic and fast procedure for the double porosity aquifer parameter determination that eliminates graphical errors inherent in the conventional TCMT. The network receives pumping test data and provides the user with the aquifer parameter values.