Methods of system identification for monitoring slowly time-varying structural systems

It is often necessary to monitor the response of structural systems for reasons such as the prediction and avoidance of unstable behavior like flutter, monitoring structural parameters (e.g. natural frequency and damping, response of critical structural modes), and detection of structural damage or verification of structural integrity. Several different system identification methods are studied for the efficacy of their implementation for online monitoring of slowly time-varying structural systems. A few simple examples of such systems facilitate the evaluation of the performance, accuracy and ease of use of the various methods. Special attention is focused on exploring H/sub /spl infin//-based identification methods and on a two-stage algorithm using a recursive least-squares identification. It is observed that H/sub /spl infin//-based methods, while having advantages for many problems, especially in computing plant uncertainty bounds that are suitable for H/sub /spl infin// control design, are not particularly well-suited for online structural monitoring. Recursive least-squares methods, however, are sufficiently computationally simple to provide real-time updates of structural parameters and, in conjunction with a Kalman filter, modal response with relatively good accuracy for systems with slowly varying characteristics.