Forecasting shear stress parameters in rectangular channels using new soft computing methods.

Area of Science:

• Hydraulics and Fluid Mechanics
• Computational Fluid Dynamics
• Environmental Engineering

Background:

• Shear stress is crucial for estimating channel velocity and discharge.
• Wall shear force percentage (%SFw) improves shear stress estimations.
• Accurate modeling of shear stress is vital for hydraulic engineering.

Purpose of the Study:

• To predict %SFw, non-dimension wall shear stress, and non-dimension bed shear stress in smooth rectangular channels.
• To compare the performance of Bayesian Regularized Neural Network (BRNN), Radial Basis Function (RBF), and Modified Structure-Radial Basis Function (MS-RBF) models.
• To evaluate the superiority of the MS-RBF model against existing equations for trapezoidal channels and rectangular ducts.

Main Methods:

• Utilized eight experimental datasets from smooth rectangular channels.
• Developed and compared BRNN, RBF, and MS-RBF models for predicting %SFw, non-dimension wall shear stress, and non-dimension bed shear stress.
• Assessed model performance using Root Mean Square Error (RMSE).

Main Results:

• The MS-RBF model achieved superior performance with RMSE values of 3.073, 0.0366, and 0.0354 for %SFw, non-dimension wall shear stress, and non-dimension bed shear stress, respectively.
• MS-RBF demonstrated better accuracy compared to BRNN and RBF models.
• The MS-RBF model outperformed three other proposed equations for trapezoidal channels and rectangular ducts.

Conclusions:

• The MS-RBF model is highly effective for predicting shear stress parameters in channels.
• This study highlights the potential of advanced machine learning models in hydraulic engineering.
• MS-RBF offers a more accurate alternative for shear stress estimation in various channel geometries.