Chloride threshold level determination: Call for test standardization to limit variations in experimental methodology and resolve inherent experimental and modelling detection challenges

Abstract

Chloride-induced corrosion significantly threatens the durability of reinforced concrete structures, leading to deterioration, costly repairs, and potential structural failures. Accurately determining the steel reinforcing bar (rebar) chloride threshold level (CTL) is crucial for predicting corrosion onset, optimizing material selection, and estimating the service life of these structures. An ensemble machine learning model was trained using literature CTL data. Despite achieving a mean absolute error of 0.218 % weight of binder, a root mean square error of 0.321 % weight of binder, and an R² value of 0.751 on unseen data, the model's performance reveals limitations due to the wide variability in reported CTL, stemming from disparities in experimental methodologies including set-up and corrosion detection techniques. After model development, this paper also investigates challenges associated with CTL evaluation by comparing literature practices and providing insights to enhance data reliability and comparability. Factors impacting CTL evaluation includes corrosion detection techniques, initiation criteria, chloride introduction methods, testing setup, exposure solution compositions, chloride concentration measurement techniques, and rebar concrete/mortar cover thickness. This paper focused on the largely ignored aspect of these factors, some of which are inherent and nearly non-circumventable, and will continue to lead to suboptimal performance of any CTL predictive model when not addressed. Recommendations for standardizing practices are proposed to improve CTL assessment consistency, reliability of developed CTL predictive models, and accuracy of service life modeling.