Model-based active noise control using neural networks

This paper presents an identification procedure of an acoustic noise model for local sound control purposes. The microphone sensor was placed near the loudspeaker, in a way that ensures robustness and stability of the plant, making possible to verify the influence of the zone of quiet in the performance of the control design. The obtained data represents a nonlinear dynamic system, due to the actuator, which led to the use of nonlinear identification techniques. Artificial neural networks already proved that could be applied in any continuous input/output mapping. The use of artificial neural networks allows the possibility of modeling acoustic fields in a simpler way, particularly for multivariable systems. A feedback control strategy was used, comparing the performance of linear and nonlinear controllers in the control of tonal disturbances.