| Real-time health monitoring on large-scaled complex structure and key engineering is the requirement of time development. Moreover, the key problem of real-time health monitoring system is structural damage identification. Damage identification method based on vibration analysis is the most prevalent issue at present. Owing to the advantages of neural network: non-linear, tolerance and robust, it is very useful to solve adverse problems, such as structural damage identification. But for the damage identification of large-scaled complex structure, as the result of the burdensome workload on picking up samples, large-scaled network and the slow training rate, this method became unpractical. Based on the studies of former scholars, damage identification of long span suspension bridge is studied by combining the damage identification method based on vibration analysis and neural network in this paper.First, damage identification method based on vibration analysis and structural damage identification by neural network are formulated in this paper, also the use of ANSYS FE analysis software in modal analysis and MATLAB neural network toolbox are introduced. And then, the feasibility of structural damage identification method by combining modal parameters and neural network is analyzed, a damage identification method using combined parameters is proposed, approaches to modeling, training and damage identification are also elaborated. On this basis, damage identification of main-cable of Yangtse River Bridge in Yangluo, Wuhan is studied. For the sake of mitigating the workload on picking up samples, downsizing the network and accelerating training, damage sections of the main-cable of the bridge are plotted. Input parameters are optimized by sensitivity analysis. Samples are picked up by modal analysis of the bridge model by ANSYS. In addition, a method for ascertaining the number of implied layer by programming using MATLAB is proposed in the paper. The BP network is trained by modified algorithm, and eventually, the optimal network model for damage identification is found out. According to the result of testing, using BP network is an effective method for the damage identification on the main-cable of long span suspension bridge. |
