Dsp implementation of active noise controller for centrifugal fans

J.-D. Wu,M. R. Bai

2014-01-01

Abstract:This paper describes an application of active noise control to reduce blade passing noise for centrifugal fans using an adaptive feedforward control structure with a control loudspeaker mounting at the cutoff region of fan is designed. The controller based on the filtered-X LMS with multi-sine reference input and frequency counting device has accelerated the convergence of the algorithm by using a digital signal processor. Experiments are carried out to evaluate the proposed design for reducing the fundamental blade passing noise and harmonic frequency noise at various flow rates. The results of experiment indicate that synthetic, periodic noise generated by loudspeakers can be reduced up to 25dB and blade passing noise from a practical centrifugal fan can be reduced up 6 to 7 dB. 1 INTRODUCTION The noise generated by fan is usually caused by the fluctuating loads on the fan blades, which results in a source of dipole order [1] and many of the sources are around the end of blades or near them. The periodic vortices can excite resonators very strongly. Most of the studies on centrifugal noise reduction were concentrated on the control of blade passing noise. The reason for the interest in this particular component of the overall noise spectrum is obvious. From a subjective point of view, the tone is generally the most annoying component and thus needs to be reduced. Many effective researches have been done to reduce the noise emitted by the fan and a lot of specialized techniques have been developed for reduce both noise sources. They have been associated with aerodynamic, and many passive silence designs have been developed for its reduction [2], [3]. In 1987, Morinushi [4] used experimental method discussed the influence of five major geometric parameters on noise and aerodynamic performance of forward curved centrifugal fans. In 1980, Neise and Koopmann [5], [6] used a quarter-wavelength acoustic resonator on the cutoff region for reducing the blade passing noise. In this paper, they indicated that the production of the tone is originated from within a concentrated region around the cutoff of the fan casing. In 1988 [7], they designed two noise control systems with control loudspeakers and vibration plates mount on the cutoff region of casing to suppress the noise using phase shifter tracking unit. In the meantime, active noise control is being advanced rapidly in the last decade due to the progress of digital signal processor technology. In these ANC configurations up to date, feedforward control has become most widely used for reduce duct noise such as fan noise when the non-acoustical reference signal is available. In the present study, a proposed application using filtered-x LMS algorithms in a centrifugal fan which showed as Figure 1 is investigated from the viewpoint of active noise control. The proposed ANC application was validated by experiments. The results showed that the system achieved blade passing noise cancellation of the centrifugal fan with a control loudspeaker mounting into the cutoff region of casing. 2 ANC SYSTEM AND CONTROLLER IMPLEMENTATION The filtered-x least mean squares (FXLMS) algorithm is an essential component of many ANC control systems. This algorithm is based on the well-known least mean squares algorithm (LMS). In this study, the FXLMS algorithm with synthetic reference is utilized in this control structure. The block diagram of Copyright SFA InterNoise 2000 2 Figure 1: The experimental setup of the duct ANC system. the FXLMS algorithm is given in Figure 2, where y (n) is the control signal, W(z ) is the transfer matrix of the FIR filter, d (n) is the primary noise, e (n) is the error signal, and the matrix Ŝ (z) represents the estimated secondary path S (z). The error signal can be expressed as e (n) = d (n)− s (n) ∗ [wT (n)x (n)] (1) where s (n) is the impulse response of secondary path S (z). Figure 2: Adaptive feedforward control algorithm with synthetic reference input. w (n) = [w0 (n) w1 (n) ......wL−1 (n) ] T is the coefficient vector of W (z) at time n, w (n) = [x (n)x (n− 1) ......x (n− L + 1) ] is the signal vector at time n, and L is the order of filter W (z). The objective of the adaptive filter is to minimize the instantaneous squared error. Assuming a mean square cost function ξ (n) = E [ e (n) ] , the adaptive filter minimizes the instantaneous squared error ξ̂ (n) = e (n) (2) using the steepest descent algorithm, which updates the coefficient vector in the negative gradient direction with step size μ. w (n + 1) = w (n)− μ 2 ∇ξ̂ (n) (3) Copyright SFA InterNoise 2000 3 where ∇ξ̂ (n) is an instantaneous estimate of the mean-square-error (MSE) gradient at time n. In practical ANC applications, s (z) is unknown and must be estimated by an additional filter Ŝ (z). The FXLMS algorithm can be written as w (n + 1) = w (n) + μx′ (n) e (n) (4) Since the fan noise contains tones at the fundamental frequency and its multiples, we use a tachometer to detect the revolution of fan. In the practical measurement, the blade passing frequency is moving lightly. The blade passing frequency is moved about in the frequency range of 6 Hz. A frequency counting technique used in DSP is to actually count the time between zero crossings to a direct measure of frequency. When a zero crossing occurs, the counter is started to increment each program cycle until the next zero crossing is encountered shown as Figure 3. The count value N is stored and the counter is reset. The blade passing frequency can be obtained from