A repair algorithm for radial basis function neural network and its application to chemical oxygen demand modeling

Area of Science:

• Artificial Intelligence
• Machine Learning
• Computational Science

Background:

• Radial Basis Function (RBF) neural networks are powerful tools for function approximation and modeling.
• Designing optimal RBF network architectures and parameters can be challenging.

Purpose of the Study:

• To present a novel repair algorithm for designing Radial Basis Function (RBF) neural networks.
• To improve the efficiency and performance of RBF networks through automated architecture and parameter optimization.

Main Methods:

• The proposed Repair RBF (RRBF) algorithm employs a two-phase approach: architecture learning and parameter adjustment.
• Architecture learning utilizes sensitivity analysis (SA) to determine optimal hidden node placement.
• Parameter adjustment refines network capabilities by modifying weights.

Main Results:

• The RRBF algorithm was successfully applied to approximating a non-linear function.
• The algorithm demonstrated efficiency in modeling chemical oxygen demand (COD) for wastewater treatment.
• Simulations confirmed the algorithm's effectiveness in both application areas.

Conclusions:

• The developed RRBF algorithm offers an efficient method for designing RBF neural networks.
• The algorithm provides a robust solution for complex approximation and modeling problems.
• RRBF shows promise for applications in environmental monitoring and data analysis.