Research On The Application Of Artificial Neural Network In Predicting The Life Of Concrete Structures

Reinforced concrete structure is widely used in industrial and civil buildings because it is combined with concrete's and steel bar's advantages, and its cost are not very high. But as the reinforced concrete buildings' service life increasing, structural performance continues deteriorating, damages accumulates step by step. These lead to the bearing capacity depresses and durability decreases. So durability assessment and life prediction for the existing reinforced concrete structures provide foundation for existing structures' maintenance, reinforcement or demolition, at the same time the research results can be directly used in durability design for concrete structures, and the study of the research is significant.The paper uses BP neural network and RBF neural network to establish the artificial neural network model for predict the depth of the concrete carbonization and the amount of reinforcement corrosion. It is divided into two-part in the contents, the first part is the research of reinforced concrete structure durability and life prediction theory. Under the condition of obtaining the measured data partially, the second part selects BP neural network and RBF neural network to fit and predict the depth of the concrete carbonization and the amount of reinforcement corrosion.The main content of the paper is as follows:(1)Concrete durability's background and research situation is introduced, and the paper analyses the concrete structural durability affected by concrete carbonization, freeze-thaw damage, alkali-aggregate reaction, chloride erosion, sulphate corrosion and reinforcement corrosion.(2)The definition of the life for concrete structures and the criterion of life evaluation are introduced. Mechanism, influencing factors and mathematics predict method of concrete carbonization and reinforcement corrosion are mainly mentioned.(3)The development of artificial neural networks and its basic theory are introduced. And the BP neural networks, RBF neural networks and their defect are mainly studied.(4)Basing on the research of reinforced concrete structure's durability and the artificial neural networks, this paper selects BP neural network and RBF neural network to predict the depth of the concrete carbonization and the amount of reinforcement corrosion, at the same time uses neural network toolbox of MATLAB to choose the network structure, transfer functions and other parameters. After lots of trial calculation and comparison between the simulation results, finally the paper proves that the prediction effect of RBF neural network is better than BP neural network, at the same time the research results can be directly used in analysis and provides foundation for structural life prediction.