Design of an Active Noise Reduction System for a Cogeneration Plant

Active noise control (ANC) aims at reducing the amplitude of a primary sound source by emitting a controlling sound wave through one or more devices, so that the two waves sum out of phase at the listening point. To generate such out-of-phase signal, the ANC system requires a specific algorithm, usually run on a Digital Signal Processing (DSP) unit. Applications of ANC systems cover a wide range of markets, from industry or vehicle interiors to consumer products, such as headphones. In this paper, a structural ANC system applied to a cogeneration plant is investigated, making use of electrodynamic shakers, accelerometers, and a microphone. The algorithm employed for generating the cancelling signal is a filtered-X Normalized Least Mean Square (FxNLMS). First, measurements of primary and secondary paths performed on the cogeneration plant are presented. Then, the off-line theoretical cancellation performance is evaluated, based on the signals’ coherence. An off-line implementation of the FxNLMS algorithm was developed and processed in several configurations, from single reference, single-input, single-output (SISO) to multiple reference, single-input, multiple-output (SIMO). Eventually, a real-time laboratory test was performed with the hardware-in-the-loop technique. The designed architecture demonstrated remarkable performance, with a noise reduction up to 4 dB(A) in the frequency range 50 Hz – 250 Hz.