Evaluating the application of data-driven intelligent methods to estimate discharge over triangular arced labyrinth weir

The use of artificial intelligence based (AI-based) methods has been known as a promising approach for solving engineering problems in order to model systems with high complexity and uncertainty such as the hydrodynamic behavior of flow passing the hydraulic structures. Considering the importance of weirs in regulating the water level and discharge controlling in water transfer channels and dams, it seems that the application of these methods can be considered as a useful tool for the estimation of discharge capacity. The present study examines the precision and use of six data-driven models including Bayesian neural network (BNN), multiple linear regression (MLR), multi-layer perceptron neural network (MLPNN), gene expression programming (GEP), least square support vector machine (LSSVM), and Chi-squared automatic interaction detector (CHAID) for the estimation of discharge passing triangular arced labyrinth weirs compared to two proposed experimental relations. To this end, 212 laboratory test results were used and statistical parameters of coefficient of determination (${\text{R}}^2$), root-mean-square error (RMSE), mean absolute error (MAE), and bias were employed as the criteria for the comparison of the models' performance. Results showed a good agreement between the observed and estimated values using the AI-based models. Among these models, the MLPNN managed to estimate the discharge passing the weir with the highest precision (RMSE = 0.00385 m³/s, R² = 0.999, and Bias<$\left|0.0001\right|$).