Deep neural network based pier scour modeling

Abstract

ABSTRACT Advancement in computing power over last decades, deep neural networks (DNNs), consisting of two or more hidden layers with large number of nodes, are being suggested as an alternate to commonly used back-propagation neural networks (BPNN). DNN are found to be flexible models with a very large number of parameters, thus making them capable of modeling complex and highly nonlinear relationships. This paper investigates the potential of a DNN to predict the local scour around bridge piers using field dataset. To update the weights and bias of DNN, an adaptive learning rate optimization algorithm was used. The dataset consists of 232 pier scour measurements, out of which a total of 154 data were used to train whereas remaining 78 data to test the created model. A correlation coefficient value of 0.962 (root mean square error = 0.296 m) was achieved by DNN in comparison to 0.937 (0.390 m) by BPNN, indicating an improved performance by DNN for scour depth perdition. Encouraging performance in present work suggests the need of further studies on the use of DNN for various applications related to water resource engineering as an alternate to much used BPNN.