Explicit Artificial Neural Networks For Predicting Gradually Varied Flow

Abstract

The ANN procedure was used to develop an explicit equation for predicting the water level profile in a gradually varied flow. The equation consists of a series of hyperbolic tangent functions, with the number of series being the same as the number on the node in the hidden layer. The ANN model consists of 3 layers: the input layer consists of four nodes, the hidden layer has seven nodes and one node in the output layer. The input parameters used are parameters related to distance, discharge, roughness, and depth of flow at the downstream end of the channel. The output parameter is the flow depth at various points. The model has been used to estimate the water level profile for different flow conditions. The comparison between the explicit ANN model and the numerical model results is satisfactory. The models can be extended to study more complex flows and non-prismatic channels. The model is promising as a tool in decision support.