An accurate estimation of the maximum possible scour depth at bridge abutments is of paramount importance in decision-making for the safe abutment foundation depth and also for the degree of scour counter-measure to be implemented against excessive scouring.Despite analysis of innumerable prototype and hydraulic model studies in the past, the scour depth prediction at the bridge abutments has remained inconclusive. This paper presents an alternative to the conventional regression model (RM) in the form of an adaptive network-based fuzzy inference system (ANFIS) modelling. The performance of ANFIS over RM and artificial neural networks (ANNs) is assessed here. It was found that the ANFIS model performed best among of these methods. The causative variables in raw form result in a more accurate prediction of the scour depth than that of their grouped form.
An estimation of scour depth is a prerequisite for the efficient foundation design of the abutment of a bridge. Many equations and models are available in literature for predicting abutment scour depth based on experimental and theoretical approaches. It is still difficult to obtain a general model to provide accurate estimation of scour depth due to the presence of complex flow structure at the base of the abutment. The artificial neural network (ANN) is generally considered as an alternative approach to the experimental and theoretical methods. It acts as a universal function approximator and consequently it is very useful in modelling problems wherein the relationship between dependent and independent variables is poorly understood. In the present study, ANN models with commonly used algorithm of training have, therefore, been developed for the equilibrium scour depth prediction at bridge abutments using a sizable amount of laboratory data from various sources. The ANN models of various training schemes have been found to be better than the conventional regression model based on the performance parameters for both calibration as well as validation set of data. The present study also indicates that the prediction based on the raw data (dimensional) is better than those based on non-dimensional parameters.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.