Abstract: Based on the corrosion experiments of reinforced concrete structure, by the use of neural network, physical models are developed to evaluate the strength of reinforced concrete structures and the content of chloride ion. To verify the accuracy of the proposed models, the content of chloride ion and the strength of reinforced concrete structures are determined experimentally. Comparison shows that the values from the physical models agree well with those of experiments..This research will help to revise the regulation on durability design of reinforced concrete structures in coastal areas.