Efficient Composite Layered Finite Element Modelling for Prestressed Concrete Box Girders

The problem of prediction of scour downstream of spillways has been mainly of an experimental nature. The traditional empirical scour prediction equations offer some guidance on the magnitude of scour parameters for a limited range. The literature review indicates that a regression mathematical model for predicting scour under all circumstances is not readily available using different flow, material and fluid parameters. The prediction of scour hole parameters like maximum depth, width and length downstream of a spillway, with empirical equations, linear regression and multilayer perceptron artificial neural networks (ANNs) has been attempted in this paper. The performance of different schemes was assessed using two error measure criteria, namely, correlation coefficient and root mean square error (RMSE). The results obtained by ANN regression are compared with conventional empirical equations. The study shows that MLP based feed forward back propagation ANN has emerged as the most satisfactory approach on the present data set as compared to the linear regression model and the empirical equations.