Artificial Intelligence for Hydraulic Engineering: Predicting discharge coefficients in trapezoidal side weirs.

Area of Science:

• Hydraulics and Fluid Mechanics
• Computational Intelligence
• Water Resources Engineering

Background:

• Accurate prediction of the discharge coefficient (Cd) is crucial for the hydraulic design and performance of side weirs.
• Traditional methods for Cd prediction can be limited in accuracy for complex geometries like trapezoidal labyrinth side weirs.

Purpose of the Study:

• To develop and compare artificial intelligence (AI) and machine learning models (MLMs) for enhanced prediction of Cd in two-cycle trapezoidal labyrinth side weirs.
• To identify the most influential hydraulic and geometric parameters affecting Cd prediction.

Main Methods:

• Three MLMs (Support Vector Machine, Artificial Neural Network, Gene Expression Programming) were developed using a laboratory dataset.
• Sensitivity analysis using the Γ-test identified five key input parameters: Fr, L/B, Le/L, (Y1-P)/P, and α.
• Model performance was evaluated using RMSE, MAE, R², and Cd(DDRmax) across training, testing, and validation phases.

Main Results:

• All three developed MLMs proved effective in predicting Cd.
• The Artificial Neural Network (ANN) model, specifically an MLP5-7-1 architecture, exhibited the highest predictive accuracy and robustness.
• The ANN model achieved excellent validation results, with RMSE = 0.0061, MAE = 0.0003, R² = 0.9301, and Cd(DDRmax) = 5.22.

Conclusions:

• Machine learning models, particularly ANN, offer a precise and efficient approach for predicting Cd in complex hydraulic structures.
• The study validates the capability of AI in advancing hydraulic engineering design and analysis.