Damage diagnosis of steel girder bridges using ambient vibration data

This paper presents an effective method for damage estimation of steel girder bridges using ambient vibration data. Modal parameters were identified from the ambient vibration data using the frequency domain decomposition technique, and were utilized as the feature vectors for damage diagnosis. Conventional back-propagation neural networks (BPNNs) were incorporated to assess damage locations and damage severities based on the modal parameters. To alleviate ill-posedness in the inverse problem, the potentially damaged members were screened using the damage indicator method based on modal strain energy (DIM-MSE). The effectiveness of the proposed method was demonstrated by means of a numerical example analysis on a simply supported bridge model with multiple girders, and by a field test on the northernmost span of the old Hannam Grand Bridge over the Han River in Seoul, Korea.