Multi-level time-variant vulnerability assessment of deteriorating bridge networks with structural condition records. (1st September 2022)

• Record Type: Journal Article
• Authors: Lei, Xiaoming; Xia, Ye; Dong, You; Sun, Limin
• Is Part Of: Engineering structures. Volume 266(2022)
• Publication Date: 2022-09-01
• Subjects: Regional bridges -- Condition assessment -- Structural inspections -- Bridge network -- Machine learning; ML Machine learning; Structural engineering -- Periodicals; Structural analysis (Engineering) -- Periodicals; Construction, Technique de la -- Périodiques; Génie parasismique -- Périodiques; Pression du vent -- Périodiques; Earthquake engineering; Structural engineering; Wind-pressure; Periodicals; 624.105
• DOI: 10.1016/j.engstruct.2022.114581
• Languages: English
• ISSNs: 0141-0296

Abstract

Highlights: Based on deep learning techniques and the Bayesian network model, an overall assessment framework of bridge networks is established. The U-Net based deterioration model established complicated nonlinear relationships of regional key features from the years of bridge inspection reports. The proposed multi-level time-variant vulnerability analysis method comprehensively assesses bridge network service performance. Abstract: Bridges are critical to ensuring access to all parts of the transportation network, and they are also the most vulnerable infrastructures. This study proposes a multi-level vulnerability assessment framework that considers structural deterioration effects and network characteristics. The condition deterioration of regional bridges is predicted with the machine learning model, which is trained with years of regional inspection data. The failure of a bridge network is converted into a Bayesian network model that takes bridge failure probability into account while assessing network vulnerability. The service performance of bridge networks is assessed with the multi-level time-variant vulnerability analysis method. The global and local network vulnerabilities are revealed through the edge-level, path-level, and network-level analysis. The proposed multi-level vulnerability assessment method is validated with a real regional bridge network. The trained U-Net model achieves high prediction performance for predicting the future condition of regional bridges. The service performance of bridge networks is comprehensively assessed with the proposed multi-level time-variant vulnerability analysis method. The corresponding results could provide a reference for the safety of managing regional bridges from a network-level viewpoint.