ABSTRACT
Wireless Sensor Networks (WSN) form a promising technology for the deployment of dense, low-cost sensor arrays for the monitoring of large-scale civil structures. The main drawback, limiting the applicability of such a solution within the industry and monitoring communities, is the restricted energy sustainability of these sensors. This problem is even more pronounced when the nodes of the network are used for the purpose of vibration monitoring, as this particular application requires both high sampling rates and excessive data transmission costs. The framework proposed in this work attempts to reduce transmission and therefore energy consumption for WSNs by employing model-based compression on output-only data. System identification procedures, such as the Eigensystem Realization Algorithm (ERA) and the Unscented Kalman Filter (UKF), are recombined in order to accurately track the underlying states of the system and extract structural characteristics, leading to continuous condition assessment and detection of possible damage. The ERA, along with an appropriate coordinate transformation procedure, are employed for assembling the structural model, whereas the UKF allows for a time-domain based reconstruction of the signal from a reduced number of sparse measurements.
