Prediction of Local Scour Depth Downstream of Sluice Gates Using Harmony Search Algorithm and Artificial Neural Networks

Abstract

Using two coupled models, this study predicts the maximum local scour depth downstream of sluice gates. The models are an artificial neural network (ANN) coupled with the harmony search (HS) algorithm, and an ANN coupled with a generalized reduced gradient (GRG) method. The models are trained and tested using extensive observations obtained from the literature. The main parameters used to predict the scour are apron length, densimetric Froude number, tailwater depth, and median sediment size. In addition, multiple linear regression (MLR) is applied to express the relationship between independent and dependent variables. Results of the ANN model coupled with HS and with GRG and of the MLR are compared. The performance of ANN is more effective when coupled with the HS algorithm. To increase the ability of the HS algorithm, a parameter varying method is applied. Results lead to the conclusion that ANN coupled with the HS algorithm is an accurate and simple method for predicting the maximum scour depth downstream of sluice gates.