Nonlinearity-tolerated active noise control using an artificial neural network

Abstract

A nonlinearity-tolerated neural active noise control scheme is presented. Time-space patterns are adaptively integrated within its architecture. A learning algorithm with time-delayed memory corresponding to the secondary acoustic paths is adopted. Simulation experiments with a hybrid structure of vibrating radiation and sound in an enclosure are conducted. It is demonstrated that the proposed approach can achieve effective noise attenuation over the whole spectrum of interest, even with a strong nonlinear environment, while the conventional filtered-x LMS active noise controller falls in chaos.