Evaluation of Artificial Intelligence Systems for Simulation of Bridge Piers Scouring in Cohesive Soils

Scouring around bridge piers as one of important and effective factors in bridge damage is a kind of erosion around the pier that occurs due to the effect of complex vortex flows and generally, result is a trench around the pier of bridge. Many field and laboratory investigations of scour around bridge pier and foundations cause to give different equations for prediction of scour depth. But, results of these equations are not satisfactory. In this research, capability of artificial intelligence is evaluated for simulation of cylindrical bridge pier scouring using laboratory data of cohesive soil bed. Using FFNN (Feed Forward Neural Network), RBF (Radial Basis Function), GRNN (Generalized Regression Neural Network) neural networks and ANFIS, scouring depth is calculated for both of dimensional and non–dimensional data and then using sensitivity analysis, effect of all parameters on pier scouring is determined. The results indicate that the ANFIS model leads to better results than the RBF and GRNN models but it is not as robust as FFNN.