Study on Damage Identification of Reinforced Concrete Members Based on BP Neural Network

Structural damage identification plays an important role in the performance evaluation and maintenance of existing structures and post-disaster damaged buildings. On the basis of summarizing and analyzing the commonly used damage models of concrete structures and the description of structural safety state in domestic and foreign codes, a method of damage index fitting and safety state identification for reinforced concrete flexural members based on BP neural network is proposed in this study. Two independent BP neural network damage models for damage index fitting and safety state recognition, respectively, are established using Matlab. For the damage index fitting, this method takes the crack characteristic parameters as the input of the network, and the damage index calculated by the dual-variable damage model based on stiffness and energy is regarded as the output. For the safety state recognition, the proposed method takes the crack characteristic parameters as the input and the safety state classification result following the FEMA-356 code serves as the output. Thereby, the mapping relationship between the crack characteristic parameters and reinforced concrete member damage can be established. Compared with the traditional damage assessment methods, this proposed method has the advantages of accuracy, promptness and convenience, and it enriches the technical means of structural health monitoring.