Bridge Damage Localization Through Response Reconstruction with Multiple BP-ANNs Under Vehicular Loading

Abstract

Damage detection is a critical aspect of bridge health monitoring. While data reconstruction has been posited as a promising method for damage detection, its effectiveness in this context has rarely been empirically validated. In this study, we introduce a novel approach to pinpoint potential bridge damage by reconstructing bridge inclination data. For an intact bridge, we selected reference cross-sections and trained multiple Backpropagation Artificial Neural Networks (BP-ANNs) to simulate transfer matrices for inclination between these base sections and other sections of the bridge. These BP-ANNs were then employed to reconstruct inclination data at the same cross-sections on a bridge with artificial damage. We demonstrated that damage localization is feasible through a comparison of the reconstructed and actual measured responses. The theoretical underpinnings of the transfer matrix and the damage localization method were initially elucidated through an analysis of the dynamics of a simplified vehicle–bridge interaction (VBI) system. A series of finite element models were constructed to substantiate the theoretical basis of the damage localization method. Additionally, a large-scale laboratory experiment was carried out to assess the practical effectiveness of the proposed method. The proposed method has been demonstrated to effectively pinpoint the location of potential structural damage. It successfully differentiates between areas in close proximity to the damage and those that are more distant. Compared to existing research, our method does not necessitate prior knowledge of factors such as mode shape functions, traffic conditions, or the constraint of inspecting with a single vehicle. This approach is anticipated to be more convenient for engineering applications, particularly in the development of online monitoring systems, due to its streamlined requirements and robust performance in identifying damage localization.

Keywords

• bridge health monitoring
• damage detection
• response reconstruction
• backpropagation artificial neural networks
• vehicle–bridge interaction system