Vibration arises in almost all moving structures. Vibration control is important to many applications such as robotic arms, aircraft wings, buildings in wind, vehicle transmission systems, to name but a few. The objective of this thesis is to develop more efficient intelligent controllers for vibration suppression, mainly for time-varying flexible structures.

At first, based on TS0 and TS1 fuzzy models, novel neural-fuzzy (NF) controllers are developed for active vibration control of the flexible structures. The NF control paradigms are intended to integrate the advantages from both fuzzy logic and neural networks while overcoming their respective limitations. The control reasoning is undertaken by fuzzy logic whereas the fuzzy control system is optimized by neural network related training algorithms. A new strategy is suggested to simplify the architectures of the classical NF controllers so as to make the control process computationally efficient for real-time applications. A recurrent identification network (RIN) is developed to adaptively identify system dynamics of the timevarying flexible structures. When system dynamics (e.g., mass, stiffness, and damping) varies, the proposed RIN and NF controller can effectively recognize the system ’s new dynamics and perform corresponding control operations. A novel hybrid training technique based on real time recurrent learning (RTRL) and least square estimate (LSE) is suggested for real-time training of the RIN scheme to optimize its nonlinear input-output mapping.

The effectiveness of the developed intelligent controllers and the related techniques has been verified by online experimental tests of corresponding fixed and time-varying dynamic conditions. Test results have shown that the developed adaptive N F controller outperforms the classical controllers (e.g., PD) and other related intelligent control strategies.