Corrosion current density prediction in reinforced concrete by imperialist competitive algorithm

Area of Science:

• Materials Science
• Civil Engineering
• Computational Intelligence

Background:

• Corrosion of steel reinforcement in concrete is a significant durability issue.
• Accurate prediction of corrosion current density is crucial for infrastructure assessment.
• Existing prediction models may lack sufficient accuracy and flexibility.

Purpose of the Study:

• To develop and evaluate an artificial neural network (ANN) model for predicting corrosion current density in concrete.
• To optimize the ANN model's performance using the imperialist competitive algorithm (ICA).
• To compare the efficacy of the ICA-optimized ANN model against a genetic algorithm (GA) for corrosion prediction.

Main Methods:

• Utilized temperature, AC resistivity (over and remote from steel), and DC resistivity as input parameters.
• Employed an artificial neural network (ANN) architecture.
• Optimized ANN weights using the imperialist competitive algorithm (ICA).
• Benchmarked the ICA-ANN model against a genetic algorithm (GA) through training, testing, and prediction phases.

Main Results:

• The ICA-ANN model exhibited enhanced ability, flexibility, and accuracy in predicting corrosion current density.
• Performance comparison indicated the superiority of the ICA-ANN approach over the GA.
• The optimized model effectively captured the complex relationships between input parameters and corrosion behavior.

Conclusions:

• The imperialist competitive algorithm effectively optimizes artificial neural networks for predicting concrete corrosion.
• The ICA-ANN model offers a more accurate and flexible solution for corrosion current density prediction.
• This approach holds promise for improved structural health monitoring and maintenance strategies in reinforced concrete structures.