Flexural strengthening of corroded steel beams with CFRP by using the end anchorage: Experimental, numerical, and machine learning methods

Abstract

This article presents the mechanical behavior of corroded steel beams that have been strengthened with carbon fiber-reinforced polymer (CFRP) layers in order to mitigate the effects of corrosion. Six beams are analyzed experimentally, including unreinforced and CFRP-reinforced specimens, with regard to the corrosion percentage, location, and shape on strength, ductility, and modes of failure. In the beam with 50 % corrosion, reinforcing with CFRP compensated for the strength reduction. In the beam with 100 % corrosion, after CFRP reinforcement, the strength was only 4 % lower than that of the control beam. A new end anchorage system was developed to avoid CFRP slippage, ensuring full utilization of its tensile capacity. Numerical modeling further validated the experimental results and then numerical specimens were used for parametric and Machine Learning (ML) studies. The results indicated that corrosion in the upper flange gave the most severe strength reduction up to 39.7 %, although this was effectively mitigated by CFRP reinforcement. The ML prediction showed that the CatBoost algorithm had the highest accuracy, with an $R^2$ score of 0.954. Additionally, the feature importance analysis revealed that the location and level of the corrosion are the most influential features affecting the reduction in the capacity of the corroded beam.