Active Noise Reduction in Ventilation Ducts

9
Proceedings of Meetings on Acoustics Volume 11, 2010 http://asa.aip.org 160th Meeting Acoustical Society of America Cancun, Mexico 15 - 19 November 2010 Session 3aSP: Signal Processing in Acoustics 3aSP8. Active Noise Reduction in Ventilation Ducts Oliver Gaab*, Delf Sachau and Oliver Pabst *Corresponding author’s address: Mechatronics, Helmut-Schmidt-University, Holstenhofweg 85, Hamburg, 22043, Hamburg, Germany, [email protected] ANC is commonly used in several noise exposed areas. The basic Idea has been extensively analyzed in literature, espe- cially for cancellation of sinusoidal noise in ducts. Here, the challenge for an ANC-System is the attenuation of undesired broadband noise in a ventilation duct with high air flow. In this case special consideration must be given towards the performance of the acoustic sensors while exposed to turbulent flow. The concept of a Swinbanks-source in combination with an adaptive filter appears to fit well for the given problem. Furthermore, causality constraints concerning the feed forward adaptive controller are analyzed with respect to sensor- and actuator placement. In an experimental setup with a duct of 8 m length and a diameter of 350 mm, noise attenuation of 14 dB in a frequency range from 50<f<450 Hertz can be achieved downstream of the control source. Published by the Acoustical Society of America through the American Institute of Physics Gaab et al. © 2010 Acoustical Society of America [DOI: 10.1121/1.3543875] Received 9 Nov 2010; published 30 Dec 2010 Proceedings of Meetings on Acoustics, Vol. 11, 055001 (2010) Page 1

Transcript of Active Noise Reduction in Ventilation Ducts

Page 1: Active Noise Reduction in Ventilation Ducts

Proceedings of Meetings on Acoustics

Volume 11, 2010 http://asa.aip.org

160th MeetingAcoustical Society of America

Cancun, Mexico 15 - 19 November 2010

Session 3aSP: Signal Processing in Acoustics

3aSP8. Active Noise Reduction in Ventilation Ducts

Oliver Gaab*, Delf Sachau and Oliver Pabst

*Corresponding author’s address: Mechatronics, Helmut-Schmidt-University, Holstenhofweg 85, Hamburg,22043, Hamburg, Germany, [email protected]

ANC is commonly used in several noise exposed areas. The basic Idea has been extensively analyzed in literature, espe-cially for cancellation of sinusoidal noise in ducts. Here, the challenge for an ANC-System is the attenuation of undesired broadband noise in a ventilation duct with high air flow. In this case special consideration must be given towards the performance of the acoustic sensors while exposed to turbulent flow. The concept of a Swinbanks-source in combinationwith an adaptive filter appears to fit well for the given problem. Furthermore, causality constraints concerning the feedforward adaptive controller are analyzed with respect to sensor- and actuator placement. In an experimental setup with aduct of 8 m length and a diameter of 350 mm, noise attenuation of 14 dB in a frequency range from 50<f<450 Hertz canbe achieved downstream of the control source.

Published by the Acoustical Society of America through the American Institute of Physics

Gaab et al.

© 2010 Acoustical Society of America [DOI: 10.1121/1.3543875]Received 9 Nov 2010; published 30 Dec 2010Proceedings of Meetings on Acoustics, Vol. 11, 055001 (2010) Page 1

Page 2: Active Noise Reduction in Ventilation Ducts

INTRODUCTION

One-dimensional active noise cancellation in ducts is often the first contact people have with the field of active noise control. Though often associated simplifications, there are however several practical problems, which can be solved with a one-dimensional approach. In contrast to noise control problems in three-dimensional space, one-dimensional noise propagation can be treated with manageable effort in terms of hardware and controller complexity. Hence this paper shows an approach to solve noise problems in ducts with the help of a boat air conditioning system. The noise caused by a large fan is of random nature in a frequency range between 50 Hz and 300 Hz. The boat fan is arranged in an acoustically isolated engine room. The ventilation system is placed at the fan output and disperses the air via a duct with a diameter of 350 mm throughout the boat cabins. Due to the rotating fan there is a noise level of about 100 dB in the duct at the fan outlet. The air velocity is about 15 m/s.

To achieve an eligible broadband sound attenuation in the duct, an ANC system as shown in FIGURE 1 is designed.

FIGURE 1. ANC-System consisting of sound source, reference and error microphone, adaptive filter and Swinbanks source.

Disturbance is detected in front of the noise source which is assumed to emit noise with an unknown spectrum. The error signal is measured downstream of the antinoise source. The delay between the correlated signals ( )x n and

( )e n allows an adaptive filter to compute a speaker output for the anti noise-source to minimize the sound pressure level. In addition to that, feedback to the reference microphone can be avoided by a Swinbanks source configuration of the secondary loudspeakers [7]. The challenge when using this method is on the one hand to design a fast and exact controller. On the other hand it is important to find proper sensor and actuator hardware and positions which can cope with the high air flow inside the duct.

delay

AF

2 m

error mic

reference mic

( )x n( )e n ( )y n

sound source

Gaab et al.

Proceedings of Meetings on Acoustics, Vol. 11, 055001 (2010) Page 2

Page 3: Active Noise Reduction in Ventilation Ducts

MOCK-UP

Hardware

The ventilation system will be represented in a mock-up made of PVC duct segments. It has a total length of eight meters and a diameter of 250 millimeters. The sound pressure sensors are 1/2” class-1 B&K microphones, type 4188. To reduce noise due to air flow turbulence at the sensor, the microphones must be geometrically adapted. This was done by using a nose cone and a turbulence screen (see FIGURE 2 and FIGURE 3).

FIGURE 2. Nose cone at B&K 4188 mic FIGURE 3. Turbulence screen for B&K 4188 mic

The primary noise is generated by a B&K “Omnisource”. The secondary speakers are typical 6” mid range woofers. The hardware providing controller unit is a dSPACE rapid prototyping system. This system consists of a Power-PC and several I/O channels. The advantage of the dSPACE platform is the easy implementation of MATLAB/SIMULINK control designs for real time applications. Signal conditioning is provided by analogue low pass filters and amplifiers.

Controller

A feed forward controller concept was chosen for this application because a reference is available with Swinbanks source. This is done by an adaptive feed forward controller concept. It has the advantage of self-adjusting the transfer function on a weak stationary reference signal. Here a FIR-filter, whose coefficients are updated by a filtered reference signal least mean square algorithm (FxLMS), is used for stability and robustness reasons [1]. According to that the FIR-filter structure delivers required stability and robustness.

Parameter

The resulting system size is 1x1x1 (reference x error x secondary source). The secondary Swinbanks sound source contains two coupled loudspeakers, which act as one unidirectional source. For controller implementation and imaging the transfer functions a SIMULINK model is developed. All additional parameters are shown below.

• Sampling rate: 4 kHz; dt = 0.25 ms • Secondary path coefficients: 512 • Controller filter coefficients: 1024 • Modeling, controlling and primary source with band-limited white noise, 0 – 1 kHz, ca. 90 dB

Gaab et al.

Proceedings of Meetings on Acoustics, Vol. 11, 055001 (2010) Page 3

Page 4: Active Noise Reduction in Ventilation Ducts

EXPERIMENTAL RESULTS

The mock-up was constructed as shown in FIGURE 4. The left border of the mock-up is moulded with an anechoic boundary. The right border is reverberant. The distance between primary and secondary source is about 2 meters. REFMIC is a microphone delivering the reference signal and ERRMIC is a microphone delivering the error signal.

FIGURE 4. ANC-System layout

The general approach for all following considerations can be separated into the following two steps:

1. Plant modeling and ANC of the feedback speaker

FIGURE 4 is a schematic illustration of the concept with the introduced Swinbanks source. It is used to avoid feedback of the secondary source to the reference microphone. Therefore the secondary path � (path between FB and REFMIC) has to be modeled. The LMS controller sets the coefficients of � to delete sound propagation of SEC to REFMIC. Hence the unidirectional secondary sound source is created. � and � are both implemented in the “FILTER”-block shown in FIGURE 5, and will be further quoted only as �.

FIGURE 5. Plant modeling and feedback suppression

Gain

Lowpass filter

Converter

Gaab et al.

Proceedings of Meetings on Acoustics, Vol. 11, 055001 (2010) Page 4

Page 5: Active Noise Reduction in Ventilation Ducts

2. Plant modeling and comprehensive ANC

As in (1.) the secondary path modeling of � is firstly done anyhow. In FIGURE 6 this is the path between secondary source and ERRMIC. Then the primary source is excited with band-limited white noise. The reference signal is acquired by REFMIC, the reference sensor. ERRMIC delivers the error signal, which will be minimized by the FxLMS. The FxLMS-algorithm now sets the filter coefficients of � and thus the output mode of the secondary source. System parameters, especially those of the adaptive controller are for instance the step size of the adaptive filter. As a result, the sound pressure level beyond ERRMIC is expected to be strongly attenuated.

FIGURE 6. Plant modeling and complete ANC-system

Performance without flow

The first test bench is done in the mock-up with a pure acoustic sound excitation. Therefore, there is no need for a special consideration of sensors and actuators in terms of flow. As it can be seen in FIGURE 7, in the current setup a reduction of the primary noise in a frequency range between 50 Hz and 850 Hz can be accomplished (measured at the ERRMIC). The grey line is the primary noise field while the solid black line provides the sound field with ANC on.

��� ��� ��� ��� ��� ��� �� �� ��� �� ���

��

��

��

��

��

��

� ����������

FIGURE 7. Results without air flow and acoustic primary excitement

Gaab et al.

Proceedings of Meetings on Acoustics, Vol. 11, 055001 (2010) Page 5

Page 6: Active Noise Reduction in Ventilation Ducts

As it can be expected, the ANC works best at low frequency < 300 Hz. The bulk of attenuation is performed there. The frequencies between 300 Hz and 600 Hz show some mock-up inherent resonance effects. As it is common knowledge, at increasing frequencies passive absorbers are more effective than an ANC-system. After this test results it is evident to prove the system with flow in the duct. This is focused in the following chapter.

Performance with flow

In this chapter the performance of the ANC-system under influence of stationary flow is investigated. For this experiment, the mock-up is arranged in an anechoic chamber and supplied with a silent airflow. The plant providing these parameters is normally used for acoustic research for aircraft ventilation systems. The primary sound source is still the omnisource, thus a pure acoustic device. The sensors are now customized for the flow conditions inside the duct. With regard to best coherence of REFMIC and ERRMIC it has been evaluated, that the turbulence screen introduced previously was the best option for sound pressure measurement. The flow was set to 8 m/s, which is lower than the conditions in the original requirement but was the maximum of the ventilation drive in the test chamber. The control performance is shown in FIGURE 8. The grey line is the sound pressure level without ANC and the black line is with ANC on. In a spectrum 50 – 600 Hz the sound attenuation amounts ca. 9 dB. That is, as expected, less than the performance without flow but still well audible. At higher frequencies the performance collapses and only marginal attenuation above 400 Hz is the result.

� ��� ��� ��� ��� ��� ��� �� �� ���

��

��

��

��

��

��� ����������

FIGURE 8. Results with silent air flow and acoustic primary excitement

Gaab et al.

Proceedings of Meetings on Acoustics, Vol. 11, 055001 (2010) Page 6

Page 7: Active Noise Reduction in Ventilation Ducts

PERFORMANCE TEST AT REAL VENTILATION SYSTEM

Customizing the hardware

For the performance test with the original fan a completely new mock-up was constructed. The test plant was positioned in a large production hall with additional sound sources. Due to limited possibilities in the given test environment, no further optimization for speaker and sensor placement was possible. Nevertheless an engineering approach for sensor and actor arrangement occurred.

The new mock-up consists of stainless steel duct segments with a diameter of 350 mm. FIGURE 9 shows the complete test bench.

FIGURE 9. Re-designed mock-up at the test bench

While reprocessing the data from the previous measurement with air flow, further customizations of the sensors and actuators were required. For better signal coherence the microphones are kept outside of the flow, enclosed in a small plastic box (FIGURE 10). These devices are connected to the inner sound field of the duct via a small slit along the length of the microphone box [1]. In order to do further decrease flow influence on the sensors the slit was closed with a special sound permeable plastic membrane. The speakers were arranged as shown in FIGURE 11.

FIGURE 10. Mock-up details and ANC-System FIGURE 11. Mock-up details and ANC-System

Fan

Intake Outlet

Speakers

REFMIC

Gaab et al.

Proceedings of Meetings on Acoustics, Vol. 11, 055001 (2010) Page 7

Page 8: Active Noise Reduction in Ventilation Ducts

Results

FIGURE 12 illustrates the performance of the ANC with a nominal air flow speed of 15 m/s. At some frequencies the controller reduces the sound pressure of the duct noise up to 10 dB. Especially at 60, 107, 136 and 220 Hz. Considering the coherence spectrum below an explanation for the results is given as the coherence of the sensor signals is high at these frequencies. For better results in a larger bandwidth, the coherence of the signals must be increased (higher than 0.9 to reach a reduction of sound pressure level by 10 dB).

� ��� ��� ��� ��� ��� ��� ���

��

��

���

��� ����������

����������� !"��#$�%�&�#������!'��(%�)�**+��,�(-�������������� !"��#$�%�&�#������!'��(%�)�**+��,�(-���

� ��� ��� ��� ��� ��� ��� ���

�.��.��.��.��.��.��.�.�.�

�/�0���,��!"��#$�1'�2#$�%�&�#������!'��(%�)�**+��,�(-���

FIGURE 12. ANC with original fan at the test bench

CONCLUSION

In this paper, the initial performance of an ANC system for ducts was shown to be functional in attenuating broadband disturbances. Future work will be focused on improving system performance and sensor / actor placement optimization, especially improving signal coherence between reference- and error microphone.

Gaab et al.

Proceedings of Meetings on Acoustics, Vol. 11, 055001 (2010) Page 8

Page 9: Active Noise Reduction in Ventilation Ducts

REFERENCES

[1] KUO, S.M., MORGAN, D.R. – Active Noise Control Systems, Algorithms and DSP Implementations, Wiley, New York, NY, 1996.

[2] ELLIOTT, S.J. – Signal Processing for Active Control, Academic Press, San Diego, CA., 2001. [3] KUTTRUFF, H. – Acoustics Primer, Hirzel Publishing, Stuttgart, Germany, 2004. [4] HONG, J., BERNSTEIN, D.S. – Bode Integral Constraints, Colocation, and Spillover in Active Noise

and Vibration Control, IEEE Transaction on Control Systems Technology, Vol. 6, NO. 1, 1998. [5] AL BASSYIOUNI, M., BALACHANDRAN, B. – Control of enclosed sound fields using spill over

schemes, Journal of Sound and Vibration 292, 2006, pp. 645-660. [6] GREßKOWSKI, J., SACHAU, D. – Comparison of active concepts for global noise reduction with

respect to implementation in aircraft cabins, 10th International Conference on Recent Advances in Structural Dynamics, Southampton, UK, 2010.

[7] SWINBANKS, M.A. – The active control of sound propagation in long ducts, Journal of Sound and Vibration 27, 1973, pp. 411-436.

Gaab et al.

Proceedings of Meetings on Acoustics, Vol. 11, 055001 (2010) Page 9