Vibration-based structural damage detection using a decentralized network with limited sensors

One of the main purposes of modal identification is to estimate the structural mechanical properties, in order to detect a possible damage over time. Several researches have been conducted recently to implement identification techniques based on the assumption of modest number of sensors. Such methods are needed both in the case of decentralized systems of wireless smart sensors, where operating units interact with each other in small groups, and in the case of repeated tests, carried out to identify large structures with a low availability of sensors. In this paper we propose a hybrid adaptive technique for natural vibration-based dynamic identification in the time-frequency domain. The proposed method involves the introduction of a novel procedure for decoupling the modal responses contained in the recorded signals, led by a clustering-based decomposition and reconstruction process; subsequently, decoupled signals, considered as single degree of freedom responses, are subjected to identification procedures in time domain, with the aim of studying the trend of instantaneous modal parameters. Since the procedure involves the real time estimation of dynamic parameters through a wavelet-based technique, it is particularly suitable for non-stationary signals. From the estimated parameters, a structural damage detection procedure is applied by using a flexibility-based technique. In order to test the effectiveness of the proposed method, the identification and damage detection procedures are simulated on a finite element model that represents a reinforced concrete plane frame.