Tracking Dynamic Characteristics of Structures Using Output-Only Recursive Combined Subspace Identification Technique

Identifying the dynamic characteristics of structures under ambient conditions is very important for heath monitoring and damage assessment. In this study, an output-only recursive combined subspace identification technique is proposed for tracking the modal parameters of a time-varying structure under unknown and unmeasured nonstationary input. The technique consists of two parts: initial input estimation and online tracking. For initial input estimation, a short-duration input is first reconstructed through an iterative oblique projection and an orthogonal projection. The reconstructed input and the measured output are then used to form a cross-correlation matrix for initial estimation of modal parameters. An orthogonal projection and an instrumental variable approach are then incorporated in order to eliminate the effects of nonstationary input and measurement noise, respectively. A bi-iteration subspace tracker is applied to extract the unknown input and the structural modal parameters. The proposed technique was validated numerically on a four-degree-of-freedom structure and experimentally on a three-degree-of-freedom building model. Both results showed that the proposed technique can simultaneously reconstruct the unknown input and track the time-varying structural modal parameters using only output measurements.