2014 Volume 70 Issue 2 Pages I_63-I_72
This paper investigates the practicability of a damage detection technique utilizing the statistical patterns of modal parameters to a real simply supported steel-truss bridge, which was consecutively subjected to five damage scenarios. The modal parameters are identified from the bridge vibrations excited by a passing vehicle via the stabilization diagram aided multivariate autoregressive analysis. The damage detection task is achieved using the Mahalanobis-Taguchi System (MTS), a multivariate pattern-recognition method. Several combinations of modal parameters as MTS variables are tested for their efficiency. Observations demonstrate that considering multiple modal frequencies as MTS variables yielded highly sensitive Mahalanobis distance (MD) to not only the presence but also the severity of the artificial damage. On the other hand, it was hard to utilize damping ratios as MTS variables for the damage detection.