Multi-approach assessment for predicting submerged discharge reduction factor in porous broad-crested weirs

Abstract

Environmentally friendly porous weirs have attracted the attention of researchers and engineers due to their favorable characteristics, surpassing solid weirs in terms of environmental impact, hydraulic performance, and stability. However, accurately estimating the submerged discharge coefficient for porous weirs is challenging due to the complex flow mechanisms involved, particularly under submerged conditions. The discharge under submerged conditions is typically expressed as a multiple of the free flow discharge, along with a coefficient representing the submerged discharge reduction factor (SDRF). This study aims to propose a novel artificial intelligence framework that incorporates metaheuristic techniques to predict SDRF for porous broad-crested weirs (PBCWs). The research utilized generalized normal distribution optimization (GNDO) to optimize the multilayer perceptron (MLP) model, enabling more precise predictions. The performance of the hybrid MLP-GNDO model was compared to that of an MLP, gene-expression programming (GEP), and standard nonlinear regression (SNR) models. A dataset comprising 966 observed experiments was employed to evaluate the proposed models. The results demonstrated that the hybrid MLP-GNDO model outperformed the MLP, GEP, and SR models, achieving a root mean square error of 0.021 and 0.022 and an R2 value of 0.964 and 0.954 for the training and test datasets, respectively. This model accurately predicted the train and test datasets with an average error rate of less than 2%. Regarding accuracy, the models ranked in the following order: MLP, GEP, and SNR.