抄録
The filtered-x LMS algorithm which needs a reference signal to provide advanced information about primary disturbances has been commonly used for active noise control in the form of the feedforward control. Therefore, for an effective feedforward control, high correlation between primary disturbance and reference signal is required. However a real system is usually out of this condition. Because there are many cases that have no space for locating the reference sensor in the system while setting the reference sensor, there may be a problem of low correlation between the primary disturbances and the reference signal. This study proposes a hybrid controller, which is an assembled form of feedforward controller and feedback controller to emphasize the strong points of the feedforward controller and overcome its faults. The feedforward controller attenuates the primary disturbances highly correlated with the reference signal, while the feedback controller cancels noises that is unobserved by the reference signal but observed by the error sensor (plant noise). The fact that the intensive periodic components of noise are commonly intended for control makes the narrowband noise canceling of the feedback controller useful beyond its limits in reduction of broadband noise. This study proved the superiority of the hybrid algorithm by making a comparison between feedforward filtered-x LMS algorithm and hybrid filtered-x LMS algorithm in the same excitation condition at the duct system of which the wave motion is a one dimensional plane wave flow. The hybrid filtered-x LMS algorithm is even more effective when the plant noise unobserved by the reference sensor is intense.
