Correlating Binder Types and Chloride Diffusion Coefficients of Concrete Mixes Using Gene Expression Programming

There is a high risk of chloride-induced corrosion of reinforced concrete (RC) in chloride-prone environment. Such corrosion of embedded rebars within concrete can drastically reduce the service life of the affected structure. Hence, evaluation of corrosion vulnerability of a RC element under vulnerable environment is necessary to ensure desired durability. The intrusion of chlorine ions in concrete occurs through the interconnected pores within concrete matrix. A high chloride diffusion coefficient of a concrete mix usually indicates a high intrusion susceptibility of chloride in the resultant RC element. One of the effective ways to reduce diffusion coefficient of concrete is through pore refinement that can be achieved by secondary hydration. The pore refinement breaks the connectivity of the pore network and eventually, hinders the intrusion of chloride ions. The secondary hydration is ensured by using supplementary cementitious materials (SCM) with active silica content. The Rapid Migration Test (RMT) is used to measure chloride diffusion coefficient of a concrete mix that requires specific test setup and preparation of test samples. The RMT is also a destructive test and time-consuming. As a result, conducting RMT experiments and finding chloride diffusion coefficients of a concrete mix is not always feasible. In this study, an attempt has been made to develop a correlation between relevant mix design parameters and chloride diffusion coefficients of the resultant concrete using Gene Expression Programming (GEP). GEP is an adaptive evolutionary artificial intelligence algorithm that uses evolutionary approaches on provided data and fits the data to generate an equation. Iterating the GEP program over a small set of data many times gives a good approximation. A total of 60 experimental RMT data has been utilized with mixes having wide ranges of SCM contents (fly ash and slag up to 40% level of replacement) and water to binder ratios. The developed correlation appeared to be quick and good estimation of diffusion coefficients that could be used as the basis for initial mix design process. Such assessment of diffusion coefficients would also assist in optimizing binder types and content for ensuring longer service life of RC structure in aggressive environment.