System identification of dynamic structural systems using continuous-time domain methods
Conte, Joel P.
Master of Science
Structural system identification is the process of deducing the properties of a structural system from its measured response to ambient vibration by fitting a mathematical model. The objectives of this study are to: (1) investigate the use of existing system identification methods, and (2) develop new system identification methods, in order to evaluate both loading and structural modal parameters of ambient excited structures for which the load process is difficult to measure. An example is the case of offshore platforms subjected to sea waves. The study considers the inverse problem from the viewpoint of continuous-time linear dynamic systems. An existing structural identification technique defined for the deterministic case (when the loading is known) and based on sequences of modal minimizations (called modal sweeps) is formulated for the general case of multiple-input multiple-output (MIMO) systems. A global minimization technique based upon the Levenberg-Marquardt algorithm for nonlinear least-squares problems is developed for the same case. Both identification techniques are applied to a multi-degree-of-freedom (MDOF) shear building model and the results are shown to be consistent. Using the framework of random vibration theory, these techniques are then converted to the stochastic case, namely when the loading process is known only statistically. These identification methods, both individually and combined, were tested based on simulated cases of increasing complexity. The results obtained are promising and indicate that under certain conditions, both load and structural parameters can be estimated from the measured response and the statistical properties of the loading process. However, the load parameters are not as well estimated as the structural parameters, since the load process is further removed from the response process than the structural filter. Although a generic shear building model has been used throughout this study to simulate the dynamic response of real structures, the results obtained are believed to apply to linear multi-degree-of-freedom systems in general.
Civil engineering; Electronics; Electrical engineering