It is important to collect reliable measured data for proper bridge health monitoring. However, in reality incomplete and unreliable data may be acquired due to sensor problems and environmental effects. In case of sensor malfunction, parts of measured data are missing and thus health monitoring cannot be carried out reliably. Due to environmental effects such as temperature variation, dynamic characteristics of natural frequencies may change as if the structure is damaged. The paper proposes a systematic procedure of data processing and data analysis for reliable structural health monitoring. Also, it applies the Mahalanobis distance as a health index computed statistically using revised data. The proposed procedure has been examined using numerically simulated data from a truss structure and then applied to a set of field data measured from Seohae cable-stayed bridge.

Reliable response measurements are extremely important for proper bridge health monitoring but incomplete and unreliable data may be acquired due to sensor problems and environmental effects. In the case of a sensor malfunction, parts of the measured data can be missing so that the structural health condition cannot be monitored reliably. This means that the dynamic characteristics of natural frequencies can change as if the structure is damaged due to environmental effects, such as temperature variations. To overcome these problems, this paper proposes a systematic procedure of data analysis to recover missing data and eliminate the environmental effects from the measured data. It also proposes a health index calculated statistically using revised data to evaluate the health condition of a bridge. The proposed method was examined using numerically simulated data with a truss structure and then applied to a set of field data measured from a cable-stayed bridge.