Advanced Disturbance Rejection Control of Smart Flexible Structures

Abstract

Controller design, as an integral step in the overall design of smart structures, plays a crucial role in active vibration suppression. Whereas well established control techniques like optimal LQG or PID controllers may perform well under assumption of the linear structural behavior, which can be described with sufficient accuracy by an LTI model, control task becomes much more complex in the presence of nonlinearities and uncertainties. In this paper we propose a feedback controller based on the recurrent wavelet neural network (RWNN) which is designed and trained to track the states of an ideal state-feedback controller, designed for the underlying linear model of the plant. In addition, adaptive neural network observer is designed to estimate the unnknown model dynamics associated with the nominal LTI model of the plant. Real time implementation of the proposed controller is realized on a Hardware-in-the-Loop setup with a flexible clamped-free beam and dSPACE system and tested for disturbance rejection tasks through a worst-case study in the presence of disturbances which cause resonant states.