Artificial neural network modeling to predict complex bridge pier scour depth

Habibeh Ghodsi, Mohammad Javad Khanjani, Ali Asghar Beheshti

Tuesday 30 june 2015

12:54 - 12:57h at Europe 1 & 2 (level 0)

Themes: (T) Water engineering, (ST) Computational methods, Poster pitches

Parallel session: Poster pitches: 5E. Engineering - Computional

Flow mechanism around a bridge pier is complicated and difficult to present a general model to provide a good prediction of scour depth. Geotechnical and economical parameters govern design of complex bridge pier design. The interaction between flow parameters and complex bridge pier is necessary to study to accurately predict the performance of system. In this study, an artificial neural network (ANN) has been developed to predict scour depth around complex bridge pier. ANN model, feed forward back propagation, FFBP, was utilized to estimate the depth of scour hole. 82 experiments have been carried out to collect experimental data. The training and testing experimental data on local scour depth around complex piers are selected from several references. Three categories of input data were used for network training: the first input combination includes cases that pile cap was above the original bed; the second combinations includes semi buried pile cap, both combinations contains 15 dimensional parameters; and the third combination includes cases that pile cap was below the original bed level which contains 8 dimensional parameters. ANN results have been compared with the results of empirical methods. Sensitivity analysis showed that the pile cap width, transversal extension of pile cap from column, and pile cap length in the first, second, and third combinations, respectively, are the most effective parameters in this process.