Application of Neural Network for Scour and Air Entrainment Prediction

This research aims at introducing a system independent method for scour and air entrainment prediction utilizing Artificial Neural Network (ANN) based on previous experimental plunge pool scour tests for inclined circular jets. Furthermore, the current manuscript introduced a single ANN model to predict air entrainment devoid of pre-knowledge of the jet condition either smooth or rough jet. Regarding ANN applicability validation, its prediction results was compared to the earlier experimental results for three regression models; one for scour, and two air-models for a smooth and rough jet. The results from each model out of the three ANN models are proved more accurate than the corresponding pre-developed regression models. Relative error envelop of 5% was found to bound all the records for the prediction of air in both ANN models (smooth and rough). For the prediction of the scour, the ANN model was also better than the regression model with only two data records of 20% relative error.