Estimating Groundwater Pollution Source Location from Observed Breakthrough Curves Using Neural Networks

This paper presents the results of a study aimed at estimating groundwater pollution source location from observed breakthrough curves using neural networks. Two different methods of presenting the breakthrough curves to the ANN are investigated. The feed-forward multi-layer perceptron (MLP) type artificial neural network (ANN) models are employed. The ANNs were trained using the back-propagation training algorithm on simulated data. A new approach for ANN training using back-propagation is employed that considers two different error statistics to prevent over-training or under-training of the ANNs. The preliminary results indicate that the ANNs are very efficient tools for estimating the distance of the potential pollution source from the observation well where breakthrough curve is measured.