Degradation law of bond strength of reinforced concrete with corrosion-induced cracks and machine learning prediction model

Abstract

During long term operation service, reinforced concrete (RC) structures are subjected to corrosive media such as chloride salts, which results in the degradation of interface bond performance, subsequently reducing the load bearing capacity and service life of the structures. To accurately assess the durability of RC structures in environments such as oceans and salt lakes, this study combines experimental methods with machine learning (ML) techniques to explore the degradation patterns and predictive models of bond performance under chloride salt corrosion. Accordingly, a series of central pull-out tests were conducted based on an electrochemical accelerated-dry-wet cycle corrosion regime to investigate the degradation patterns of bond performance between corroded steel bars and concrete under different corrosion degree and corrosion-induced crack widths. By comparatively analyzing the evolution patterns of cracks and the morphology of the steel bars in reinforced concrete specimens and their failure modes, and integrating the mechanical degradation mechanism with chemical and microscopic corrosion mechanisms, an in-depth analysis of the corrosion-induced degradation patterns at the bond interface was performed. The internal relationships between corrosion degree, corrosion-induced crack width and bond strength were discussed, and the variation patterns of bond strength with corrosion-induced crack width and corrosion degree were explored. Moreover, the bond strength patterns were analyzed from an energy perspective. The results indicate that bond strength and bond energy decrease with an increase in the corrosion-induced crack width of the concrete. Additionally, based on the experimental data from this study, ML predictive models for bond strength were established using corrosion-induced crack width as the control variable. The results show that the ML-based bond strength predictive models fit well with the experimental data, offering higher accuracy and reliability compared to traditional bond strength models. The research findings provide significant theoretical basis and practical guidance for safety assessment, maintenance, reinforcement, and full-life design of corroded RC structures.