Prediction of the Groundwater Rebound Process in a Backfilled Open Cut Mine Using an Artificial Neural Network

Abstract

A neural network model was used to predict the groundwater rebound process after cessation of dewatering at a restored open cut coal site in the East Midlands area of the UK. Time (days after dewatering), water table levels in the aquifer and the backfilled site, hydraulic conductivity of the aquifer and backfilled site, and precipitation were used as input. The output of the network was the water table height, until the water table reached its equilibrium position. A feed-forward artificial neural network that uses batch gradient descent with a momentum-learning algorithm and 6-1-6-1 arrangement was found capable of predicting the groundwater rebound process. Predicted values were very close to the monitored results. The correlation coefficient values were 0.98221 for the training set, and 0.99329, 0.99499, 0.98667, 0.98289, and 0.97141 during the testing stage for the five monitoring points, showing that the model prediction was satisfactory.