ニューラルネットワークを用いたコンクリート構造物中の鉄筋の腐食進行予測

抄録

Corrosion of reinforcement in concrete structures is one of the serious deterioration. Because the crack or the reduction of cross section of reinforcement by corrosion causes the deterioration of theperformance of structures or the functional disorder of structures. In this study, we discuss the method for evaluating the state of corrosion in the actual structures based on the condition of current structures and environmental condition. Then we assume that the carbonation depth is not only the index of pH value at bar location but also the index of oxygen supply at bar location. The model for evaluating the state of corrosion is constructed as the neural network trained by the data from site investigation. The effect of the factors on the progress of corrosion is discussed by the network, The network indicated quantitatively that the condition for occurrence of corrosion and the condition for the corrosion to be severely. In addition, it is clarify that the progress of corrosion is possible to be predicted by the combination of the neural network, prediction method for the carbonation and chloride movement referred from past studies.