2022 Volume 78 Issue 1 Pages 11-22
This study proposes a damage detection method for bridges using a damage indicator that is automatically derived from vehicle induced vibrations. To extract damage-sensitive features, the principal component analysis is applied to the characteristic polynomials of auto-regressive models estimated from measured acceleration. A stochastic distance between damage-sensitive features from the healthy bridge and unknown test samples evaluates anomalies involved in the modal properties. A statistical hypothesis testing based on the stochastic distance was performed for damage detection. A field experiment on an actual steel truss bridge whose truss members were artificially severed was conducted to investigate efficacy of the proposed method. Observations showed the proposed method enables to detect three different damage patterns. Efficacy of the proposed method was also demonstrated by comparing to the previously investigated damage-sensitive features.
