Multi-Mode Control of Structures by Using Neural Networks with Marquardt Algorithms

Abstract

This paper presents an integrated approach of identification, estimation and vibration control of smart structures by using neural networks with Marquardt adaptation algorithm. An identification neural network is first constructed to model the system dynamics of a composite laminate structure embedded with piezoelectric sensor and actuator. Based on the identification network, a neural controller is then developed to meet the vibration suppression performance specified by a reference model. In addition, an estimation neural network is also developed that enables the measurement of a single piezoelectric sensor to represent the state variables of displacement and velocity of the structure. The neural controller can then minimize the structure vibration by using a pair of piezoelectric sensors and actuators. Experimental verification shows that the vibration amplitude of the smart structure under two-mode harmonic excitation and random excitation can be reduced by 80% and 60%, respectively.