This study presents a development to the wave method for use in structural damage detection and damage localization in bridges. The method was shown to be robust when applied to real structures and large amplitude response in buildings. A companion research using a uniform shear beam model showed opportunities in using the wave method for system identification purposes in bridges. However, the shear beam model was shown to be too simple for capturing the change at the component level (i.e., columns). Therefore, a more detailed model was proposed which consists of a uniform Timoshenko beam (TB) model. One main advantage of the TB compared to the shear beam is its ability to account for flexural (bending) deformation of the structure. This is particularly important since the bending motion can cause significant wave dispersion (i.e., wave travel with different speeds at different frequencies of motion) which is not captured by a simple shear beam. The identification process in this study includes estimating the shear (cS) and the longitudinal (cL) wave velocities for the structure by fitting an equivalent uniform TB model in impulse response functions of the recorded acceleration response. The identification process is further enhanced by adding the model’s damping ratio to unknown parameters. Hence, for each pair of recorded acceleration response, the identification algorithm finds three unknown variables for the best fitted TB model. Next, the structural damage is detected by detecting changes in the identified longitudinal velocities from one damaging event to another.

In this study, the acceleration response from a shake-table tested bridge were used. The bridge was tested with 7 biaxial motions which progressively damaged the bridge to failure. Availability of a shake-table test data provided us an opportunity to assess the accuracy of our identification algorithm and to determine if damage in the bridge can be localized. The complexity of the bridge response led us to create a detailed nonlinear finite element model (FEM) of the bridge for further investigation of its response. The FEM was subjected to the acceleration response recorded at the shake tables. The FEM was further updated to match the observed response of the bridge. Similar identification algorithm was applied to the response extracted from the FE model. A summary of damage observed at the actual bridge was collected during each shaking intensity. For the FEM, the hinge formation at the columns during the seven shakings were monitored and compared. A comprehensive comparison between the reduction in identified longitudinal velocities and the actual observed damages were presented. The results revealed that the TB was accurate in capturing overall change in the bridge due to damage. For each column, the TB model showed more change in column 1 relative to other columns which was consistent with the observed extent of damage in the bridge. While the uniform TB model has shown significant improvement compared to a simple shear beam model, further study will be needed to assess the goodness of a layered TB model with varying properties along its length. Such layered TB can potentially account for contrast in the stiffness at the deck and cap-beam level and the stiffness of the columns along the wave propagation passage.