| Inferring useful information from observations is what identification concerns. The primary mission of a structure health monitoring (SHM) system is to provide valuable insights into the current condition and the future behavior of a structure, through analyzing the acquired data. The achievement of the mission requires the development of advanced identification techniques, for civil structures are usually quite complicated and the measurements from a SHM system tend to be heavily contaminated by measurement noise.;Although the approaches within the maximum likelihood framework have dominated the identification community since the 1970s, it still suffers from two major deficiencies. First, an appropriate parametric model needs to be established, which is a quite cumbersome task. Second, the estimates obtained by iterative optimization may get stuck into local minimum far from the true solution. Due to the deficiencies, these approaches are particularly difficult to be implemented for complex civil structures. Subspace identification, on the other hand, circumventing the two problems through exploring the subspace structure of a system and making use of some robust numerical algorithms such as the singular value decomposition (SVD), seems fairly promising for civil applications. In this research, several novel subspace identification techniques are proposed for SHM. As to system identification, we propose an approach to estimate structural parameters with unknown earthquake excitation using only absolute acceleration responses. The unknown earthquake excitation can be recovered simultaneously. In the method, the system is assumed to be linear time invariant (LTI). Identification of linear time-varying systems in ambient condition is next investigated. Besides system identification, we also develop a subspace identification technique to track noise variance which may be with time-varying characteristics. Some numerical examples and real tests are used to illustrate these techniques Results show that these techniques exhibits remarkable performance and have a great potential for practical applications... |
