Artificial Neural Network in Bridge Monitoring Application

• Main Authors: Tseng,Yao-Hsiarn, 曾耀賢
• Other Authors: Chang,Kuo-Chun
• Format: Others
• Language: zh-TW
• Published: 1994
• Online Access: http://ndltd.ncl.edu.tw/handle/73827168955004984515

碩士 === 國立臺灣大學 === 土木工程研究所 === 82 === Identifying patterns of changes in vibrational signatures of a bridge structure is a promising approach for on-line structure monitoring.Artificial neural networks,developed by researchers in cognitive sciences and artificial intelligence, is a tool that can used to classify patterns according to previously learned experience. The main benefit of using neural network model is its ability to classify signals that are fuzzy or imprecise. In this thesis,studies on the application of neural network system to a bridge monitoring system is described. Artificial neural network (ANN) is employed to detect the changes of vibrational signatures of a two-span three-girder bridge model with various types of simulated damage. In this approach, these different types of signature changes are used as these state variables of the monitoring system, and are used as the input for the net. Damage locations of the bridge are predicted by the output of the net.Training examples are randomly generated, and the net is trained by these training examples to achieve the learning purpose. After the net is trained, the damage locations of the bridge structure can be deduced from the input of the measured displacement mode shape of the damaged structure. Numerical simulations show that application of artificial neural networks (ANN) in analyzing the change of patterns in displacement mode shape of bridge structures has considerable potential to structural damage diagnosis and condition monitoring. In summary, it can be concluded that an ANN may be developed as an efficient tool for damage detection of a bridge monitoring system.