Ultra-high precise Stack-LSTM-CNN model of temperature-induced deflection of a cable-stayed bridge for detecting bridge state driven by monitoring data. (November 2022)

• Record Type: Journal Article
• Authors: Yue, Zixiang; Ding, Youliang; Zhao, Hanwei; Wang, Zhiwen
• Is Part Of: Structures. Volume 45(2022)
• Page Start: 110
• Page End: 125
• Publication Date: 2022-11
• Subjects: Cable-stayed bridge -- Temperature-induced deflection -- Data driven -- LSTM -- CNN -- Abnormal detection; Structural engineering -- Periodicals; 624.1
• DOI: 10.1016/j.istruc.2022.09.011
• Languages: English
• ISSNs: 2352-0124

Abstract

Abstract: To deduce the bridge state through the deflection of the main girder of a cable stayed bridge, the mapping relationship between temperature features and temperature-induced deflection is modelled by deep learning. Through mechanical mechanism, adequate and logical temperature information is extracted. By the data-mechanism dual-driven mode and the advantages of combing Long Short-term Memory (LSTM) with Convolutional Neural Network (CNN), the improved Stack-LSTM-CNN with higher interpretability and reliability is used for modelling. Residual value between the measured value and the regression value from mapping model is used for indicating abnormity. Benefiting by the high precision of the Stack-LSTM-CNN, only processing the residual by moving average, the abnormal deflection can be detected with the sensitivity during 5–17 mm. This precision shows that the improved Stack-LSTM-CNN model has the potential to detect damaged cable based on bridge geometry, which can't be achieved by multiple regression or existing neural networks.