Author(s): Isa Ebtehaj; Azadeh Gholami; Saeed Reza Khodashenas; Hossein Bonakdari

Linked Author(s): Hossein Bonakdari

Keywords: Bank profile shape; Experimental model; Group Method of Data Handling (GMDH) model; River morphology; Threshold channel

Abstract: This paper presents a study on the design of threshold channels with a sand bed, focusing on the natural shape of bank profiles in addition to the width and depth of the water surface. The research involves extensive laboratory studies on steady-state channels and the development of a neural network model based on the Group Method of Data Handling (GMDH). The GMDH structure is optimized with a genetic algorithm, and laboratory data from four discharges are used for calibration and validation. The performance of the developed GMDH model is compared to eight models proposed by other researchers, including statistical and analytical studies, as well as a nonlinear regression model. Comparing the results from the ten models against the laboratory values demonstrated that the GMDH-GA model displayed the highest accuracy and simplicity simultaneously in the test phase (MARE = 0.05551; RMSE = = 0.0719; R = 0.9705; AIC = -89.12) when compared to the theoretical- and the artificial intelligence and entropy-based models. The GMDH-GA model proposed in this research suggests a polynomial curve for the cross-sectional shape of the proposed threshold, which was deemed the most suitable model among the other proposed shapes when compared against the laboratory values. Furthermore, the GMDH model provides a simple and practical relationship for predicting the dimensions of the cross-section in all types of channels, which can be applied in the design, construction, and implementation of artificial channels.

DOI: https://doi.org/10.3850/978-90-833476-1-5_iahr40wc-p0984-cd

Year: 2023