Prediction of Water Discharge Through the Foundation of a Rockfill Dam in Brazil

Abstract

In this work, water discharge through the soil foundation of Corumbá-I rockfill dam is predicted using artificial neural networks and the Box and Jenkins autoregressive approach for time series. The analysis is carried out based on a sequence of water discharges measured through the residual soil foundation near the left dam abutment, during the period between 28/08/1997 and 30/12/2002. Application of neural network techniques for prediction and analysis of data from geotechnical instrumentation may be an invaluable tool for performance monitoring of dams and other civil engineering works. In the case of Corumbá-I dam, the artificial neural networks yielded quite satisfactory results with respect to water discharge behavior, including the dependence of the predicted responses on several other factors such as the reservoir water level and the pore pressure values.