Active Noise Control Using Neural Network

Abstract

An active noise control technique has been developed using piezoelectric transducers as an actuator and a neural network as a controller. In this study, the fully coupled acoustic-plate interaction equations including the characteristics of piezoelectric transducers are solved using time-dependent Green's function techniques, where a time-varying mean air density is considered. The control system introduced here consists of a forward model network for structural identification and a controller network, both of which are two-layered networks having linear transfer functions. The performance of the controller is discussed experimentally as well as numerically in a model which consisits of a simply supported plate in an infinite rigid baffle above a rectangular cavity.