Application of enhanced hybrid optimization models for discharge prediction of cosine sharp-crested weirs

The discharge coefficient plays a crucial role in estimation of discharge over sharp-crested weirs. This study employs various hybrid optimization algorithms to predict the discharge coefficient over half-cycle and full-cycle cosine sharp-crested weirs. A Support Vector Machine (SVM) algorithm was utilized, with its parameters optimized using the Gray Wolf Optimization (GWO) algorithm. The performance of the GWO-SVM hybrid model was compared against the Gaussian Process Regression (GPR) model. Additionally, Gene Expression Programming (GEP) was applied to derive the best predictive equations for both weir types. For this purpose, a broad range of laboratory data, comprising geometric and hydraulic information (110 data sets for half-cycle and 270 data sets for full-cycle cosine weirs), has been considered. The results indicated that the GWO-SVM model demonstrated high accuracy, particularly in peak discharge predictions, achieving errors below 4 %. The proposed models incorporated the influence of the approach Froude number, a critical factor often overlooked in prior discharge prediction models. Sensitivity analysis revealed that the ratio of weir height to upstream water head (h_o/P) and Froude number play significant roles in prediction of discharge coefficient for cosine sharp-crested weirs. The frequency histogram of prediction for half- and full-cycle cosine sharp-crested weirs showed that the GEP model exhibited a uniform error distribution with a slight underprediction tendency, whereas regression models were less reliable for both weir types. Overall, the AI-based models outperformed conventional regression approaches, effectively minimizing underpredictions and overpredictions, and providing robust discharge coefficient predictions.