Structural Damage Alarming by Novelty Detection Technique Based on Modal Flexibility

摘要

Structural damage alarming is explored by using novelty detection technique. Based on the changes of modal flexibilities, a novelty index to alarm damage is generated by a feed-forward BP neural network. A total of 12 potential damage cases in a cable-stayed bridge are simulated. Modal flexibilities are evaluated using noisy 'measured' modal parameters. The results show that the novelty index based on flexibility has a high sensitivity to damage and a good tolerance to the noise. In this technique, no structural model is needed. Only a few of low-order modal parameters at several locations are to be measured.