Development of a novel narrowband active noise controller in presence of sensor error

Narrowband active noise control (ANC) systems are used to mitigate periodic or nearly periodic noises produced by large scale cutting machines, engines, turbines, compressors etc. These systems employ a non acoustic sensor such as accelerometer for identifying the frequencies present in the input noise signal. These frequency values are used by the cosine wave generator to synthesize the input waveform. However, when the frequencies sensed by the tachometer are different from the actual frequencies it results in the problem of frequency mismatch (FM). In this paper through simulations degradation in the performance of conventional narrowband ANC system is observed even in the presence of FM of about 1 %. Hence, a new method using functional link artificial neural network (FLANN) is proposed which outperforms the conventional method in the presence of less to high FM. Simulation results demonstrate the efficacy of the proposed ANC system.