Data-Driven Estimation Models for Chloride Ion Diffusion Coefficient in Concrete Using Bayesian Approach

The corrosion of reinforcing steel bars induced by chloride has a significant impact on the performance of reinforced concrete (RC) structures. Although various models have been proposed to estimate chloride diffusivity, many empirical models remain unavailable due to database limitations or the oversight of certain factors. To enhance the accuracy of chloride diffusivity estimation, this paper has collected 109 sets of experimental data. These data were used to develop a Bayesian prediction model for a more precise estimation of chloride diffusivity. The input features examined in this study include water–cement (W/C) ratio, thickness of concrete specimens (M (cm)), volume fraction of coarse aggregate (R), ratio of environmental temperature to standard curing temperature (T/T₀), ratio of environmental humidity to standard curing humidity (h/h_c), and exposure time. The output parameter is the chloride diffusion coefficient. Sensitivity analysis of the input parameters reveals that the ratio of exposure time to curing time (t/t₀) and T/T₀ are the key factors influencing the chloride diffusion coefficient, with importance coefficients of 0.83 and 0.67, respectively. The h/h_c has the least impact. In addition, the W/C ratio and R also have a certain influence on the chloride diffusion coefficient, and optimizing these parameters can further enhance the durability of concrete structures.