It is expected that damages of transportation system including highway bridges cause serious interference of traffic function and recovery efforts, when a large earthquake occurs to a densely populated urban area as it was happened during the great Hanshin Awaji Earthquake in 1995. Therefore, the real-time earthquake damage detection system for bridge structures seems to be necessary to provide information of damages in transportation network system just after the occurrence of the earthquake and to assist the damage reduction, prevention of damage expansion and recovery works. This research proposes the method of evaluating the damages of bridge pier to provide useful information for real-time earthquake damage detection system by using Neural Networks technique. In this study, the emulator was constructed first to express intact structure by using Neural Networks. Then, it was examined whether the difference between output of the emulator and measured response of the bridge pier showed the damages of the bridge structure. Dynamic experimental data which were acquired at the shaking table test of RC single column conducted by Public Works Research Institute, Tsukuba, were used to verify the proposed method. Amplitude ratio and phase difference were selected to show the difference between output of the emulator and measured response of the bridge pier. As a result, it was confirmed that these two indices corresponded well to the damages, namely the change in the natural frequencies and the damping constant. The proposed method is able to provide the useful information for damages more quickly than frequency domain analysis method.
View full abstract