Application of Meta Models for Simlulation of Bridge Pier Scouring in Non-Cohesive Soils

Scour around bridge pier is one of the most influential and important factors in the destruction of bridges. This phone man occurs due to contact and separate lines of flow of pier and complex vortex flows. Because it depends on many influential variables, determination of influential variables and formulation of mathematical models is difficult. Conventionally, the multiple linear regression procedure has been known as the most popular model in simulating the bridge pier scour. However, when the nonlinear phenomenon is significant, the multiple linear will fail to develop an appropriate predictive model. Recently, Data Mining and Meta Model approaches such as artificial neural network (ANN) and neuro-fuzzy methods have been used successfully for the bridge pier scour modeling. In this research, capability of Data Mining and Meta Model is evaluated for simulation of bridge pier scouring using laboratory and field data of cylindrical and square pier and non-cohesive soil. In one hand, using Feed Forward Neural Network, Radial Basis Function, generalized regression neural network, Adaptive Neuro-Fuzzy Inference System and experimental equations, and in other, the dimensional and non-dimensional data, scour is calculated and then using sensitivity analysis, effect of all parameters on pier scouring is determined. The results of simulation indicate that the FFNN has higher coefficient of determination and lower mean absolute relative error compared to RBF, GRNN and ANFIS models. Finally, sensitivity analysis shows that flow velocity has the greatest scour depth with dimensional data, while non-dimensional bed shear stress has the biggest influence on dimensionless data. A conclusion is reached that the Data Mining and Meta modeling approach is suitable for modeling of bridge pier scour.