ABSTRACT
Damage detection stands as a pivotal subject in bridge health monitoring. As the actual structures and their theoretical models inevitably have differences in structural parameters, it is practically challenging to detect damages in the absence of precise baseline information. In this regard, this paper introduces a baseline-free damage detection framework that combines measurements of moving loads with acceleration responses. Specifically, the moving loads are extracted using computer vision techniques while the acceleration is measured by distributed sensing. For each measured moving load, the same loading process is applied to the finite element models with different structural parameters and ambient noise. With simulated structural responses, an advanced 1d convolutional neural network is proposed for damage detection. Theoretical derivation reveals the proposed framework is applicable for damage detection without initial structural parameters. The proposed framework is validated to be effective on both damage localization and quantification.
