Effectiveness Characteristics of the Neural Network as Active Noise Control Algorithm against Residential Ventilation Openings

In this study, Active Noise Control (ANC) is applied to reduce middle-frequency and low-frequency noise entering through ventilation openings to residential indoor. The past numerical simulation by authors against road traffic noise establish that ANC effectiveness increases with applying the Neural Network (NN) as a control algorithm in comparison with the Least Mean Square (LMS) that is an orthodox algorithm. It is guessed that such effectiveness is caused by a non-linear regression function involved in the NN. It isn't explained yet, however, what characteristics of the interference noise generated by the NN are effective. Therefore, comparative analyses of control effectiveness characteristics with the NN and the LMS are performed in this paper. And then the advantages of the NN are considered quantitatively with numerical simulation on the assumption that ANC technique is applied against a circular opening. It is demonstrated as a result that the NN is about 15 dB better than the LMS on the whole of target noise. It is also demonstrated on comparison of frequency characteristics of control effects that the NN performs about 20 dB larger effects at the frequency range from 200 Hz to 800 Hz in which the noise level is large.