Damage Identification of Bridge Based on Multi-branch Convolutional Neural Network under Moving Load
Abstract
To address the challenge of detecting damage in a large number of in-service small and medium-span bridges, this study proposes a damage identification method based on a multi-branch convolutional neural network (CNN) under moving loads. Two CNN architectures—a dual-branch model and a multi-branch model—are developed for structural damage identification. The sensitivity of damage identification to sensor location is also investigated. First, various damage scenarios are simulated using a finite element model of a bridge. The structure is excited by a moving vehicle load, and the resulting structural vibration responses are extracted through transient analysis. These responses are then used as input to train and validate the established CNN models. Finally, the effectiveness and accuracy of the proposed method are verified through a laboratory-scale model test. The results demonstrate that both the dual-branch and multi-branch CNN models exhibit higher computational efficiency and better identification performance than a single-branch CNN model under multiple damage scenarios. Furthermore, the identification results show no obvious difference among sensors placed at different locations for the same damage case.
Keywords:
damage identification, convolutional neural network, moving load, finite element analysis, bridge damagePublished Online
How to Cite
Issue
Section
License
- Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (see The Effect of Open Access). As soon as the paper is accepted, finally submitted and edited, the paper will appear in the "OnlineFirst" page of the journal, thus from this point no other internet-based publication is necessary.

