Optimization of cable tension in large-span cable-stayed bridges based on RBF neural network and improved sea-gull...

Area of Science:

• Structural Engineering
• Computational Mechanics
• Reliability Engineering

Background:

• Large-span cable-stayed bridges require precise cable force optimization for structural integrity and reliability.
• Existing optimization methods may not adequately capture complex nonlinearities and reliability considerations.

Purpose of the Study:

• To develop a novel force optimization model for large-span cable-stayed bridges incorporating reliability indicators.
• To enhance the accuracy and efficiency of structural response prediction and optimization algorithms.

Main Methods:

• Established a structural surrogate model using Radial Basis Function Neural Network (RBFNN) for nonlinear mapping.
• Improved the standard Seagull Optimization Algorithm (SOA) with refracted backpropagation and nonlinear convergence.
• Integrated RBFNN and the enhanced SOA for a combined force optimization approach.

Main Results:

• RBFNN effectively modeled the relationship between structural random variables and dynamic responses.
• The enhanced SOA demonstrated superior performance over the standard SOA for optimization.
• Optimization reduced main beam deflection by up to 36.21% and improved tension reliability indicators by up to 9%.

Conclusions:

• The proposed RBFNN-enhanced SOA method is effective for reliable force optimization in large-span cable-stayed bridges.
• The approach significantly improves structural performance, particularly in reducing mid-span deflection.
• The study validates the integration of reliability indicators into the cable force optimization process.