Abstract:
This study proposes an artificial neural network (ANN) model to solve an inverse parameter identification problem for groundwater modeling. It is a problem for which the transmissivities can be obtained for given hydraulic heads. ANN may be a useful tool for parameter estimation problems because of its ability to model complex nonlinear relationships between state variables and system parameters without a priori assumptions of the nature of a relationship like a black box. To carry out a parameter estimation using the ANN, a hypothetical example has been examined under two scenarios, one involving the sink and/or source terms, the second without these. In the ANN model, the network is trained for about 5, 10, and 20% of all data, and then transmissivities in the other cells are forecasted. Results show that observed and forecasted transmissivities are in good agreement when about 10 and 20% of the hydraulic heads in the solution domain are known. Copyright © 2006 Begell House, Inc.