Hybrid Generalized Regularized Extreme Learning Machine Through Gradient-Based Optimizer Model for Self-Cleansing...

Area of Science:

• Environmental Engineering
• Computational Fluid Dynamics
• Hydraulic Engineering

Background:

• Sediment transport modeling is crucial for managing open channel sedimentation and associated operational expenses.
• Existing models often suffer from limited data ranges, impacting their accuracy and reliability in channel design.
• Accurate flow velocity computation is key for robust engineering solutions in open channel design.

Purpose of the Study:

• To develop and evaluate advanced sediment transport models using extensive experimental data.
• To compare the performance of hybrid machine learning algorithms against existing regression models.
• To assess the impact of incorporating channel parameters on model accuracy.

Main Methods:

• Implementation of Extreme Learning Machine (ELM) and Generalized Regularized Extreme Learning Machine (GRELM) algorithms.
• Hybridization of ELM and GRELM with Particle Swarm Optimization (PSO) and Gradient-Based Optimizer (GBO).
• Validation using a comprehensive dataset covering a wide range of hydraulic properties.

Main Results:

• The GRELM-GBO model demonstrated superior accuracy, outperforming standalone ELM, GRELM, GRELM-PSO, and traditional regression models.
• Models incorporating channel parameters showed increased robustness.
• GRELM-GBO achieved 18.5% higher mean accuracy compared to the best-performing regression model.

Conclusions:

• The GRELM-GBO hybrid model offers a highly accurate and robust solution for sediment transport modeling in open channels.
• The study highlights the importance of utilizing extensive datasets and incorporating relevant parameters for reliable model development.
• The findings support the practical application of advanced ELM-based methods in hydraulic engineering and environmental problem-solving.