System identification for controlled structures in civil engineering application: Algorithm development and experimental verification

The research work documented in this dissertation is inspired by the recent advances of feedback control in civil structural systems, which aims at reducing hazardous effects of large dynamic loads on the structure, like strong earthquake and high wind. The first step in these investigations is usually the development of a mathematical model that faithfully describes the dynamic behavior of the physical system. This is the major topic of our discussion.; Throughout this research, an application-oriented design strategy has been followed closely. Instead of aiming at some “universal” algorithms, we studied the physical system first, then developed the algorithms accordingly. In this regard, we have identified two classes of representative systems: linear structures controlled by either linear active damping devices or nonlinear semi-active damping devices; both are popular configurations in civil structural control.; For the first class of systems, the main result was the development of linear frequency domain identification techniques that directly attacked the problems induced by the dynamic characteristics of such systems, e.g. the lightly damped nature of the structure and the fixed zero dynamics in the control input channels. The algorithms were subsequently incorporated into a closed-loop identification scheme, and a strategy for modal-based LQG control redesign was proposed. From a different perspective on control development, we also invented a PID-like design approach that adapted to several application-driven demands.; For the second class of systems, our focus was placed on the development of nonlinear blackbox identification techniques for the modeling of magnetorheological fluid dampers. Two algorithms have been proposed in this respect: the wavelets based construction approach and the ridgenet-based optimization approach. Finally, by combining the linear and nonlinear identification techniques, a scheme was proposed to develop an integrated model for structures controlled by nonlinear semi-active damping devices.; Besides algorithm development, a significant amount of attention was put into experimental verification. The proposed algorithms were supported by carefully implemented numerical softwares and detailed application results on test facilities available in the Structural Dynamics and Control/Earthquake Engineering Laboratory at the University of Notre Dame.