USE OF ARTIFICIAL NEURAL NETWORKS IN THE PREDICTION OFLOCAL SCOUR

摘要

The primary purpose of the study was to assess the utility of artificial neural networks in theprediction of local scour. Secondary objectives were to evaluate various configurations of the artificialneural network model and to ascertain the most significant variables with respect to scour prediction. Thedata used in the study were taken from the published results of an experimental research program onscour downstream from a sluice gate. On the basis of this work, it is concluded that an artificial neuralnetwork model can be successfully used to predict local scour. It is shown that the model complexitymust be increased if variables experiencing a large range are to be successfully evaluated, and that agood model will incorporate data from the entire data domain. It is also found that the most importantparameters affecting scour depth are discharge, tailwater depth, angle of repose and porosity.