Wavelet packet transform applied to active noise control system for mixed noise in nonlinear environment

Abstract

The noise emitted from rotating machinery usually presents a mixture of broadband and narrowband frequency components. Hybrid functional link artificial neural network (HFLANN) system consisting of narrowband ANC (NANC) subsystem, broadband ANC (BANC) subsystem and sinusoidal noise canceller (SNC) subsystem is effective to suppress the mixed noise. However, in the conventional HFLANN system, the attenuation performance of broadband subsystem cannot meet the requirement, the narrowband subsystem cannot adapt to the nonlinear environment. To address the problems, an improved HFLANN (IHFLANN) system is proposed in this paper. By applying the wavelet packet (WP) structure, the broadband noise is decomposed into different frequency bands, which improves the attenuation performance of broadband subsystem. The FLANN structure is applied to the NANC subsystem to improve the steady-state performance in nonlinear environment. Meanwhile, to reduce the computational complexity, we introduce a simplified IHFLANN (SIHFLANN) system. Simulation results demonstrate the proposed systems achieve better attenuation performance and the simplified system reduces computational complexity without sacrificing the attenuation performance.