Corrosion Behavior and Corrosion Prediction of Carbon Steel under Dynamic Atmospheric Corrosion Environment in Harbin

A dynamic atmospheric corrosion test was carried out with a dedicated test vehicle operating in Harbin, China. Through corrosion kinetic analysis and corrosion product composition analysis, together with electrochemical tests, the corrosion damage behavior of carbon steel after specific periods of exposure was investigated. The corrosion rate of carbon steel showed a gradual decrease with the increase of corrosion time; the rust layer resistance R_f and charge transfer resistance R_ct gradually increased due to the hindering effect of the dense rust layer and deposition of SiO₂. Moreover, three hybrid machine learning models, including ABC-SVR, GA-SVR, and PSO-SVR, were constructed to predict the dynamic atmospheric corrosion rate. The results showed that the PSO-SVR algorithm outperforms the GA-SVR and ABC-SVR algorithms, with MAPE = 6.13%, RMSE = 1.11 μm/year, and R² = 0.9810.