Advanced prediction of the sinking speed of open caissons based on the spatial-temporal characteristics of multivariate structural stress data. (October 2022)

• Record Type: Journal Article
• Authors: Dong, Xuechao; Guo, Mingwei; Wang, Shuilin
• Is Part Of: Applied ocean research. Volume 127(2022)
• Publication Date: 2022-10
• Subjects: Open caisson -- Sinking speed prediction -- 3D-CNN -- Deep learning -- Changtai Yangtze river bridge project; Ocean engineering -- Periodicals; 620.416205
• DOI: 10.1016/j.apor.2022.103330
• Languages: English
• ISSNs: 0141-1187

Abstract

Abstract: This work explores the feasibility of predicting the sinking speed of open caissons using a convolutional neural network (CNN), which has significance to the safety and reliability of open caisson sinking. This research focuses on the relationship between the sinking speed and the multivariate structural stress data of the open caisson during the sinking process by earth excavation. A three-dimensional CNN (3D-CNN) model was proposed to predict the sinking speed in advance by extracting the spatial-temporal characteristics of multivariate structural stress data. Next, the prediction accuracy of single-step prediction, multistep prediction, and real-time prediction were verified by the actual sinking speed monitored in the Changtai Yangtze River Bridge Project. In addition, the influence of both the advantages of the proposed model and the need to extract the spatial-temporal characteristics of stress were further analysed. The results indicate that the proposed model has high prediction accuracy than two-dimensional CNN (2D-CNN) and artificial neural network (ANN) models. The prediction accuracy does not decay with an increase in prediction length in multistep prediction, showing that the proposed model is effective and practical for predicting the sinking speed using a 3D-CNN based on the spatial-temporal characteristics of structural stress.