Vibration-based structural damage detection by deep learning method on a small steel bridge structure
With the rapid development of computer science and the enhancement of the computing power of hardware,deep learning method shows its enormous advantages in dealing with large amount of data and classification issues.In civil field,structural damage detection is still a challenging topic.As structural health diagnosis can be viewed as a classification task,deep learning method may give hints and potential solution to overcome the difficulties in structural damage detection.This paper shows a new attempt to apply deep learning method on a short span steel bridge structure.Accelerometers are applied for the structural vibration data acquisition.The structural damages are simulated by applying additional element on the bridge structure to change the local structural states.A convolutional neural network(CNN)is designed to extract structural features and identify the damage locations of the steel bridge.Without hand-crafted feature extraction data processing before training,acceleration data is used directly as the training data even though the data is a little noisy.The performance of the damage detection of the deep neural network is evaluated based on the difference between the label of training data and the output of the deep neural network.The feasibility of applying deep learning method to the vibration-based structural damage detection is discussed.However,only simulated structural damage is discussed in this study.Thus,the performance of the detection of actual structural damages and the correlation between the simulated structural damage and actual structural damage will be investigated in the future.
Deep learning convolutional neural network damage detection steel bridge structural health monitoring
Y.Zhang Y.Miyamori S.Mikami T.Saito
Department of Cold Regions,Environmental and Energy Engineering,Kitami Institute of Technology,Hokka Department of Civil and Environmental Engineering,Kitami Institute of Technology,Hokkaido,Japan
国际会议
The 7th World Conference on Structural Control and Monitoring(7WCSCM)(第七届结构控制与监测世界大会)
青岛
英文
2385-2396
2018-07-22（万方平台首次上网日期，不代表论文的发表时间）