Résumé :
|
The research goal was to investigate the application of artificial neural networks (ANNs) for subsurface drainage system simulation. Specifically, back-propagation ANNs were trained to imitate a conventional mathematical model, DRAINMOD, in the simulation of water-table depths. The objectives were to analyze and discuss the impact of various strategies of data input into the ANNs. The strategies, including feedback and time lag procedures, were developed for a good representation of the dynamics of the soil system. The feedback procedure consisted of feeding the previous ANN outputs back into the current ANN inputs. The time lag procedure is correlated with feeding the input values of previous time steps. The results show that the use of feedback and time lag procedures can produce significant impacts on the ANN performances.
|