Statistical damage identification for bridges using ambient vibration data

The inherent uncertainties in experimental data have been recognized as one of the main barriers against the application of vibration-based damage identification techniques on real-life bridges. A statistical damage identification procedure for bridge health monitoring is presented in this paper. It is assumed that the structure, in both healthy and unknown conditions, is monitored and the dynamic responses under ambient excitations are available. The damage identification procedure runs following a 4-step scheme including (1) data sample formation, (2) data normalization, (3) damage feature extraction, and (4) statistical damage evaluation. A hierarchical sequence matching scheme is suggested for data normalization to account for the effects of various environmental and operational conditions on the structural dynamics. The damage feature extraction technique based on time series analysis combining auto-regressive and auto-regressive with eXogenous inputs prediction models is adopted. A statistical index based on the damage features that are derived from a large number of data samples is proposed for novelty detection and damage localization. The effectiveness and robustness of the proposed procedure is demonstrated by numerical simulations performed on a three-span continuous girder bridge with reasonable damage severity.