Corrosion of rebar in concrete. Part III: Artificial Neural Network analysis of chloride threshold data
Yakun Zhu, Digby D. Macdonald, Jie Qiu, Mirna Urquidi‐Macdonald
Abstract
Active corrosion of carbon steel in reinforced concrete occurs when the chloride ion concentration exceeds the chloride threshold (CT). CT is affected by many physical and chemical parameters including primary variables (pH, corrosion potential, breakdown potential, and temperature) and secondary variables (cemenet composition, concreete porosity, and water/cement ratio). A Kohonen-self organized map (KSOM) and regression artificial neural network (ANN) coupled methodology was developed to find the missing values of independent variables in the sparse database and for quantitatively evaluating the effects of these variables on CT values that are expressed in %TotalCl/cem (or %FreeCl/cem), and [Cl−]/[OH−].