Predicting Sediment transport in sewers using integrative harmony search-ANN model and factor analysis

Abstract

This study evaluates the performance of an integrated version of artificial neural network namely HS-ANN (which is a combination of neural network and heuristic harmony search algorithm) as an alternative approach to predict the sediment transport in terms of sediment volumetric concentration (Cv) in sewer pipe systems. To overcome the complexities of choosing the optimum number of the input variables and to consider the effective parameters of the model, the factor analysis technique is utilized. In addition to the HS-ANN model, an empirical equation, as well as a multiple linear regression model, are also considered. The mean square error (RMSE), mean absolute percentage error (MAPE), and Pearson correlation coefficients (PCC) are used for evaluating the accuracy of the applied models. As the comparisons demonstrate, the HS-ANN model (PCC = 0.97) is more accurate than the existing empirical equation and MLR model and could be successfully employed in predicting sediment transport in sewer networks.