Adaptive neurocontrollers for vibration suppression of nonlinear and time varying structures

Neural network based control systems with online adaptation are capable of coping with system uncertainty, nonlinearity, and variations with time. The current article presents a comparison of neurocontrollers with a standard LQR control system for vibration reduction. Controller performances are tested using an experimental setup employing a cantilevered plate with surface bonded PZT actuators. The large accelerations due to simisoidal (first and second mode) disturbances cause geometric nonlinearity. Time variations of system dynamics (by adding mass or attaching modification plate) and external excitations are investigated. The results are presented in both time and frequency domains and the measurement uncertainties are identified. The results show that neural adaptive predictive controller is very promising in terms of control effectiveness and control effort in the vibration suppression of nonlinear, time varying smart structures.