Experiments for damage detection by subspace identification on a tied arch bridge

If structures of civil engineering are planned for a finite life time, it is advisable to control the buildings for failures continuous such as by a monitoring system. In this lecture it is proposed to identify the dynamic characteristics of a structure by vibration measurements. In the case of damage this characteristic will be alternated. Here Black-Box state space models are used to detect and localise system variations. Black-Box State space models identified by the measurement data represent the transfer function between input and output and can be used for damage detection. Additional we will present in this lecture a transition from black-box to white-box state-space models. White-box models are physically interpretable and permit direct damage localization. Theoretical it can be shown the possibility to extract mechanical properties like mass, stiffness or damping direct from identified state space models. Because of differences between theory and experiment this is more difficult on real measurements. Nevertheless the identified model parameters (e.g. Markov-Blocks) are able to detect and localize variances of mechanical properties. Results from experimental measurements in our laboratory on a bending beam will show that the methods are able to localize damages. Furthermore experiments on a prestressed concrete tied arch bridge in Hunxe (Germany) will be presented. The bridge had a span of 62.5 meters and was deconstructed. Preliminary numerous experiments were accomplished. Presented are experimental results from a cut-through hanger at one tied arch as an induced damage.