Prediction of backwater level of bridge constriction using an artificial neural network

Abstract

Bridge constriction in channels usually increases the water level well above the normal depth and may result in overflow on the surrounding floodplain. In this paper, the experimental backwater level at which the maximum afflux value was observed due to bridge constriction was investigated. An artificial neural network (ANN) was used to predict the backwater level based on Manning's roughness coefficient of the main channel (nmc) and of the floodplain (nfp), bridge width (b) and flow discharge (Q). A multi-layer perceptron (MLP) ANN was used to predict the backwater level using these parameters. Multiple linear (MLR) regression and multiple non-linear regression (MNLR) were used as benchmarks for comparison of ANN results. It is concluded that an ANN can very accurately predict the backwater level. The developed ANN model was then used to conduct a parametric study to investigate the influence of nmc, nfp, b and Q on the backwater level due to a bridge constriction without piers. It is concluded that nmc and Q have a more profound effect on the backwater level than does nfp, while b has very little effect on the backwater level within this range of parameters. Other observations and conclusions are also drawn.