Time-varying system identification using a hybrid blind source separation method

Abstract

System identification and damage detection are one of the major building blocks of structural health monitoring for critical infrastructure. Using measured vibration response of structure, the unknown structural parameters can be extracted by any traditional system identification approach. However when there is a damage, these parameters vary with time which makes the signal nonstationary. In this paper, a simple yet effective method is proposed under the framework of blind source separation (BSS) method to identify the instant and severity of damage. Several time-frequency decompositions including Hilbert transform and time-varying auto-regressive modeling are explored to improve source separation capability of the BSS method.