Temperature effect on the modal frequencies of a steel railway bridge

This paper presents the dynamic identification of the Ostiglia-Revere railway bridge, a steel truss girder bridge located in Northern Italy. The bridge is 6.6 m wide, about 940 m long and composed of 12 spans. The accelerations of a bridge span caused by both ambient excitation and train passages have been continuously measured from August to November. The monitoring system consists of 4 temperature sensors and 4 biaxial MEMS accelerometers, acquiring accelerations with a sampling frequency of 80 Hz. Modal properties are estimated adopting two different identification approaches, namely the Enhanced Frequency Domain Decomposition and the Stochastic Subspace Identification. Particular attention is paid to two crucial issues for vibration-based structural health monitoring: the recognition of the same structural mode from results identified during different time windows (i.e.; mode clustering), and the temperature effect on estimated modal properties. As the latter is concerned, two regressive models, namely a linear regression model and an ARX model, are fitted to the frequency-temperature data, strongly reducing the eventuality of false vibration-based damage detections in the future.