Stacked ensemble model for optimized prediction of triangular side orifice discharge coefficient
Mohamed Kamel Elshaarawy, Abdelrahman Kamal Hamed
Abstract
This research focuses on optimizing the prediction of discharge coefficient (Cd) of triangular side orifices (TSO) using a novel stacked model (SM) incorporating five machine learning models (MLMs): Support Vector Regression (SVR), Gene Expression Programming (GEP), Random Forest (RF), Adaptive Boosting (Adaboost), and eXtreme Gradient Boosting (XGBoost). An Artificial Neural Network (ANN) was employed as the SM to enhance Cd prediction accuracy. The MLMs were developed using 570 experimental datasets, considering five non-dimensional parameters: TSO’s crest height to height ratio (W*), main channel width to TSO’s base length (L*), width to TSO’s height (H*), upstream flow depth to TSO height (Y*), and upstream Froude number (Fr). The SM-ANN achieved the best performance with an R² of 0.998 and a root-mean-square-error of 1.85 × 10−3, followed by XGBoost and GEP. Sensitivity analysis showed that W* and Fr had the greatest effect on Cd. A graphical interface was developed for direct Cd prediction.