Enhancing Hydraulic Performance of Labyrinth Weirs: A Comparative Analysis of GEP, ANN, and KNN Algorithms

Authors : Tohid Omidpour Alavian, Mahdi Majedi-asl, Nazila Kardaan, Mahdi Soltani Sotobadi

Pages : 69-85

View : 49 | Download : 112

Publication Date : 2025-12-28

Article Type : Research Paper

Abstract :Labyrinth weirs, as advanced hydraulic structures, play a pivotal role in managing flood flows and enhancing dam discharge capacity due to their unique periodic geometry. However, their complex design demands precise hydraulic analysis. This study evaluates the performance of Gene Expression Programming (GEP), Artificial Neural Networks (ANN), and K-Nearest Neighbors (KNN) algorithms in predicting discharge coefficients (C_D) using 243 experimental data series, incorporating geometric and hydraulic parameters such as the total head-to-height ratio (H_t/P), cycle arc angle (θ), and sidewall angle (α). Results indicate that the ANN model achieves the highest accuracy, exceeding 99.66% (R2 = 0.9966, DC = 0.9965, RMSE = 0.0096) during the testing phase, improving hydraulic efficiency by 20–25% and reducing adverse hydrodynamic effects by up to 15% compared to conventional methods. The KNN model, with a prediction error below 0.15% (RMSE = 0.0015, R2 = 0.9932, DC = 0.9933), optimizes flow by 15–18% and mitigates deviations by up to 12%. Conversely, GEP exhibits a 12–14% generalizability decline and a 116.3% error increase (RMSE = 0.0584, DC = 0.8389), limiting its efficacy by 25–30% in complex flow simulations. Sensitivity analysis identifies H_t/P as a critical parameter, influencing accuracy by 30–35%. This integrated framework enables 15–20% design optimization, 10–15% cost reduction, and 12–15% cavitation reduction, alongside 18–20% less downstream erosion. Surpassing limitations of prior empirical (e.g., Johnson, 1965) and numerical (e.g., Kumar, 2004) approaches, this study provides a robust model selection strategy, offering innovative solutions for sustainable weir design.
Keywords : Labyrinth Weirs, Gene Expression Programming, Artificial Neural Networks, K-Nearest Neighbors, Hydraulic Efficiency

ORIGINAL ARTICLE URL