Principal Component Analysis of Deflection Responses for Damage Identification in Railway Steel Truss Bridges

Abstract Many railway steel truss bridges in service today are reaching, or have exceeded, their intended design lifespan. However, these bridges are subjected to loads significantly higher than those initially considered during their original design due to the increasing needs of transportation, which may result in structural deterioration. As a result, timely damage identification and efficient maintenance are now crucial. This study proposes a deflection response-based damage detection framework for railway steel truss bridges. Vertical dynamic deflection response data are acquired for both undamaged and damaged bridge conditions and used to compute the covariance matrix from which principal components are extracted. To measure the effectiveness of the presented approach, a numerical study is conducted using a finite element model of a steel truss bridge. Damage detection and localization involve comparing the principal component positional differences of intact and damaged states. The baseline for maximum operational speed is derived using allowable speed data as a common baseline, and further analysis is carried out under various damage scenarios with added artificial noise to simulate measurement uncertainties. The obtained results show that the deflection-based PCA technique successfully identifies and localizes the zone of damage with stiffness reduction as low as 10% under single and multiple-damage scenarios, and it reliably distinguishes damage severity across a range of operating conditions. This approach provides railway authorities with a simple and scalable early damage detection tool, which may help in prioritizing inspection and improving safety management of aging steel bridges.