Design and Experimental Validation of an Adaptive Neurocontroller for Vibration Suppression

Abstract

The real time sensing and actuation capabilities of smart structures are employed for active vibration control. A neural network based adaptive controller is designed and validated experimentally for nonlinear vibration suppression. Piezoelectric actuators are used to reduce the vibrations of a cantilevered plate subject to impulse, sine wave, and band-limited white noise disturbances. A multilayer perceptron with a single hidden layer of neurons is used as the controller. For an adaptive control system, the neurocontroller learns online in real time starting with random weights and biases. For fast learning, the Levernberg-Marquardt backpropagation algorithm is implemented by writing a C-file S-function used with MATLAB/Simulink. The learning rate is varied during the minimization of the cost function. Experimental results demonstrate quick learning, excellent closed-loop performance, and robustness of the adaptive neurocontroller.