Investigation of chloride diffusion in cement mortar via statistical learning theory

Salt damage is frequently found in coastal structures and is known to be one of the leading causes of concrete degradation. To ensure a concrete structure with adequate resistance against such a threat, chloride ion diffusion in concrete needs to be thoroughly investigated. Although chloride diffusion in concrete has attracted significant attention from researchers, the method to determine the chloride ion concentration remains the subject of debate. In this study, a series of tests was first conducted on cement mortars less than 168 d old. A model based on statistical learning theory and adopting the least-squares support vector machine was developed. Comparison with the experimental results revealed that the model could provide an accurate prediction of chloride ion concentration. Moreover, the proposed model, unlike the conventional Fick's law approach, is applicable to cement mortars of differing compositions and ages, thus contributing to its applicability to practical engineering. The test variables included age, depth (measured position), dimensions of diffusion and the presence of a reinforcement bar. All the specimens were mixed with fresh water and cured in salt water.