ABSTRACT
Wireless Sensor Networks (WSNs) performing vibration-based monitoring are subject to heavy data transmission costs affecting their performance. The work presented herein addresses this issue by utilizing the spectro-temporal information contained in measured signals for recovery from incomplete spectro-temporal measurements. This in turns allows for reliable identification of modal shapes and natural frequencies. To this end, this work overviews the process of time-series recovery from the partially transmitted spectro-temporal information and provides case studies illustrating the ability of the proposed scheme to efficiently detect modal shapes from experimentally obtained operation data. The validation is carried out on an experiment setup of a 4-storey steel frame structure tested in using a laboratory shaking table. The extracted data serve for feeding the proposed data compression framework and the system identification routines. The identification results are next quantified in terms of modal shape and frequency extraction accuracy.
