ABSTRACT
Chloride ingress through diffusion is a major deterioration mechanism, making it absolutely necessary to account for it in the Service Life Prediction (SLP) framework. However, in most SLP frameworks, there are limited provisions for the estimation of the chloride diffusion coefficient (Dchloride ), a parameter that has a direct strong influence on the chloride induced corrosion initiation. Developing a probabilistic model for Dchloride based on raw material characteristics and mixture proportions can help in making decisions on the selection of cementitious materials during the planning and design phases of projects. This thesis presents the development of probabilistic models for the estimation of Dchloride based on 42 data sets collected from literature. First, the predictor variables such w/b, specific surface area of binder (SSAbinder), and SiO2 and CaO contents in the binder were identified. Then, a diagnostic study was conducted to investigate the nature and degree of influence of each independent variable on Dchloride . Then, probabilistic models that relate Dchloride to the significant independent variables were developed.
