Scour depth prediction at bridge piers by Anfis approach

Abstract

Accurate and reliable prediction of scour depth around bridge piers is important due to the complexity of the scour process. In this study, an adaptive neuro-fuzzy inference system (Anfis) approach is used to predict the scour depth around circular bridge piers. In particular, the applicability of the Anfis method as a prediction model for scour depth is investigated. A total of 165 data records are used to predict equilibrium scour depth from various experimental studies. Two different models are constructed for the prediction. The first comprises a combination of dimensional data, whereas the second includes non-dimensional input variables. The performance of the Anfis models in training and testing sets is compared with observations. The models are also compared with a radial basis neural network (RBNN), existing scour depth equations and multiple linear regression (MLR). The results of the Anfis models, RBNN, MLR and existing scour depth equations are all compared to yield a more reliable evaluation. The results show that the Anfis method can provide high accuracy and reliability for the prediction of scour depth around circular bridge piers.