Finite Element–based Data-Driven Method to Detect Multiple Damages of 1D Beam Model and 2D Slab Model of Bridges-a Theoretical and Experimental Study
Zhen Yang,Cheng Wang,Yi Li,Kang Gao,Ye Xia,Wei Gao,Gang Wu
DOI: https://doi.org/10.1016/j.measurement.2024.115709
IF: 5.6
2025-01-01
Measurement
Abstract:Bridges, being vital structural components in infrastructure with a substantial stock, are confronted with severe service conditions and serious issues of deterioration. An urgent demand exists for the application of inspection methods that are rapid, automated, and require minimal manpower to achieve a comprehensive detection of bridge damages. The present study proposes a non-destructive damage detection framework for multiple damages of 1D beam bridges and 2D slab bridges based on data-driven method from in-house finite element programing. Through finite element calculations of the displacement time history data at the vehicle-bridge contact point when a vehicle passes over a bridge simulated with multiple damage conditions, which is also the Apparent Profile (AP) data, the difference between APs when a vehicle passes over a bridge in damaged and intact state can be calculated to obtain Contact Point Displacement Variations (CPDVs). Then, Multiple damage conditions will serve as the outputs and CPDV (Contact Point Displacement Variation) under different damage conditions will severe as inputs of BP (Back Propagation) neural network. The architecture of BP neural network is then adjusted to fit the mapping relationship between CPDVs and damage parameters. Subsequently, random damages are introduced within the bridge structure. By collecting the vehicle’s acceleration time series data as it passes over the damaged bridge, AP can be calculated using only the vehicle dynamics model. Further, CPDV can be derived and used as input into a Backpropagation (BP) neural network for bridge damage detection. Finally, a scaled model experiment was carried out to verify the validity and robustness of the framework. The results revealed that the proposed method can predict bridge damages precisely and robustly, even for damages located near bridge supports of minor extent. The proposed method addresses the limitations of prior research by enhancing the detection of slab bridge damages and precisely mapping local bridge damage information. It is applicable for routine condition assessments of medium and small span bridges, providing detailed information on the location and severity of localized damages.