Toward the accurate estimation of elliptical side orifice discharge coefficient applying two rigorous kernel-based...

Area of Science:

• Hydraulics and Fluid Mechanics
• Computational Intelligence
• Data Science

Background:

• Accurate prediction of discharge coefficients is crucial for hydraulic engineering.
• Traditional methods may lack precision for complex geometries like elliptical side orifices.
• Machine learning offers advanced data-driven approaches for hydraulic modeling.

Purpose of the Study:

• To precisely estimate the discharge coefficient of elliptical side orifices in rectangular channels.
• To compare the performance of kernel-based machine learning models (GPR, KELM) with traditional methods (GRNN, RSM).
• To develop a practical equation for discharge coefficient prediction and analyze parameter sensitivity.

Main Methods:

• Utilized Gaussian Process Regression (GPR) and Kernel Extreme Learning Machine (KELM) for modeling.
• Employed Generalized Regression Neural Network (GRNN) and Response Surface Methodology (RSM) as validated schemes.
• Developed models using 588 laboratory data points under varying geometric and hydraulic conditions.

Main Results:

• Machine learning models demonstrated superior performance over regression-based relationships.
• GPR and KELM models achieved high accuracy, with GPR showing RMSE = 0.0081 and R = 0.958.
• A new practical equation was derived from the RSM model, and channel width to orifice height ratio (B/b) was identified as the most significant parameter.

Conclusions:

• Kernel-based machine learning models, particularly GPR and KELM, provide highly accurate estimations for elliptical side orifice discharge coefficients.
• The study successfully identified key parameters influencing discharge and developed a practical predictive equation.
• The approach is effective for outlier detection and defining the applicability domain in hydraulic modeling.