Damage identification of a full-scale five-girder bridge using time-series analysis of vibration data

Abstract

This paper presents damage identification of a full-scale, five-girder bridge subjected to controlled levels of sequential damage using time-series analysis of vibration data measured during field tests and also simulated by finite-element (FE) modeling. Three-dimensional (3-D) FE models of the bridge corresponding to various damage scenarios considered for the experiment are constructed and calibrated using the modal parameters of the tested bridge extracted from vibration records obtained using a relatively dense measurement grid. A time series-based damage identification technique, autoregressive with exogenous input (ARX) models and sensor clustering, is evaluated using the real bridge data and the simulated bridge data obtained from the boundary value problem. Modification of the technique through incorporation of a new damage-sensitive feature (DSF) is proposed to enhance identification of the induced damage to this highly redundant bridge. The effects of damage locations, damage extents, and vibration measurement noise on the damage identification results of the bridge are investigated. The analysis results indicate that the damage identification technique can be successfully used to identify the existence of damage in this highly indeterminate bridge. It is found that damage location, damage extent, and vibration measurement noise significantly impact interpretation of the inferred damage location. The effects of interpretation parameters on the damage identification results are discussed in detail.