Simultaneous Assessment of Damage and Unknown Input for Large Structural Systems by UKF-UI

Abstract

Much progress has been made in the assessment of structural damage and unknown input (UI) using incomplete and noisy measurement signals. The unscented Kalman filter (UKF) has proved to be a sophisticated approach to this task. A novel method using UKF with unknown input (UKF-UI) for recursive identification of a state-input system has been proposed by the authors. However, the purpose of this study was to propose the new UKF-UI framework and validate it with some simple structures. Although very limited research has been conducted on the UKF for health assessment of large structural systems, including two-dimensional (2D) and three-dimensional (3D) frame structures, it is based on a two-stage approach and requires full measurement of all acceleration, velocity, and displacement responses in the substructure containing the UI. Some implementations either have limitations in real-time identification or need assumptions on the time evolution of UI. One example is the random walk hypothesis, which heavily depends on the tuning of noise parameters. The application of UKF to large structural systems is still a challenging problem. This observation has prompted the authors to investigate the UKF-UI framework for identification of large structural systems. Here, it is extended to the assessment of damage and UI by the UKF-UI method for 2D and a 3D finite-element (FE) frame models. By the partially measured noise-polluted structural acceleration and displacement responses, the extent and location of damage is assessed at the element level. The unknown external excitations are simultaneously identified with no assumptions about the time evolutions of a one-stage identification process.