Damage Assessment with Ambient Vibration Data Using a Novel Time Series Analysis Methodology

In this study, a novel approach using a modified time series analysis methodology is used to detect and locate structural changes by using ambient vibration data. In addition, it is shown that the level of the damage feature gives important information about the relative change of the damage severity, although direct damage quantification is not achieved. In this methodology, random decrement (RD) is used to obtain pseudofree response data from the ambient vibration time histories. Autoregressive models with exogenous input (ARX models) are created for different sensor clusters by using the pseudofree response of the structure. The output of each sensor in a cluster is used as an input to the ARX model to predict the output of the reference channel of that sensor cluster. After creating ARX models for the healthy structure for each sensor cluster, these models are used for predicting the data from the damaged structure. The difference between the fit ratios is used as the damage feature. The methodology is first applied to experimental ambient vibration data from a steel grid structure, in which different damage scenarios, such as local stiffness loss and boundary condition change, are simulated. The results show that damage was detected and located successfully for most of these cases. Moreover, it is observed that the relative extent of the damage is also estimated by using the method. Then, output-only data from the Z24 bridge is used for further verification of the methodology with real-life data where different levels of pier settlement were applied as damage. It is shown that the approach is successful in damage identification and localization with a minimum number of false alarms. The potential and advantages of the methodology are discussed on the basis of the experimental results. Limitations of the approach are also addressed, along with future research directions.