Optimisation of an active control system to reduce the exhaust noise radiated by a small generator

Optimisation of an active control systemto reduce the exhaust noise radiated by a

small generator

MarõÂ a Cuesta *, Pedro CoboInstituto de AcuÂstica, CSIC, Serrano 144, 28006 Madrid, Spain

Received 4 January 2000; received in revised form 8 June 2000; accepted 5 July 2000

Abstract

In a previous paper (Cuesta M, Cobo P. Active control of the exhaust noise radiated by anenclosed generator. Applied Acoustics 2000;61(1):83±94) the authors reported a passive/activesystem to control the exhaust noise radiated by a small generator. Passive control was a�or-ded by a steel rectangular enclosure lined with a layer of absorbing material. The enclosure,

designed to provide the higher Insertion Loss, supplied attenuation higher than 20 dB above500 Hz. To reduce the noise below 400 Hz, an active control system was designed with onereference, one error input and one control output. Whilst many harmonics in this frequency

band were attenuated, the ANC system was unable to reduce some of them. The aim of thispaper is to identify the origin of this lack of attenuation and to alleviate it. An exhaustiveanalysis of both the transfer function between the secondary source and the error sensor (the

plant) and the reference signal has been carried out. A new ANC system with improved per-formance is described which controls all the harmonics until 300 Hz. # 2001 Elsevier ScienceLtd. All rights reserved.

1. Introduction

An engine is a typical noise source radiating multiple tones at the fundamental®ring frequency. Acoustical enclosures constitute the most frequently used measuresto reduce the noise radiated by engines [2]. The performance of acoustical enclosuresis based upon the principle of trapping the noise radiated by the sound source, bymassive, impervious layers, and dissipating the retained sound energy, by poroussound-absorbing lining. The e�ciency of this passive technique depends on the ratio

Applied Acoustics 62 (2001) 513±526

www.elsevier.com/locate/apacoust

0003-682X/01/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved.

PI I : S0003-682X(00 )00060 -8

* Corresponding author.

E-mail address: [email protected] (M. Cuesta).

of the sound wavelength to the thickness of the blocking and absorbing layers. Thelarger is the sound wavelength (the lower is the frequency), the thicker must be thewalls of the enclosure. Co-generation plants in hospitals and generators to supplyelectrical power to mobile TV units are examples of heavy, bulky enclosuresdesigned to provide high attenuation at the whole frequency range.More often, space is limited and enclosures are close-®tting and leaky [2]. Close-

®tting enclosures are those in which most of the enclosed volume is occupied by thenoise source. Enclosed engines must have openings for air intake and gas exhaust.Sound leaks through these openings, especially at low frequencies. It is well knownthat traditional passive techniques are less e�ective at lower frequencies.Active noise control (ANC) techniques rely on the principle of destructive inter-

ference between the existing, or primary, noise and the anti-phase, or secondary, noiseset up by an electronic controller [3]. The simpler adaptive ANC system processes areference signal to generate the control signal that drives the secondary source. Anerror signal monitors the control performance. The adaptive ®lter must extrapolatethe acoustic ®eld from the reference sensor to the error sensor (primary path) andcompensate for the transfer function between the control source and the error sensor(secondary path) [4,5]. Since ANC works best on low frequency noises, the activeapproach complements the traditional passive control methods.A hybrid passive/active system to control the noise radiated by a small generator

was reported by Cobo et al [6]. Passive control was provided by enclosing the gen-erator within a rectangular, steel box, lined with absorbing materials. Low frequencynoise escaped from the enclosure through air intake, air refreshing, and gas exhaustopenings. Exhaust noise dominated the noise radiated by the enclosed generator.The exhaust noise spectrum consisted on a series of single tones harmonically related,whose fundamental frequency, concerned with the ®ring frequency of the engine,varied slightly depending on the electrical load connected to the generator.A SISO ANC system was designed to attenuate the periodic exhaust noise. The

reference signal was supplied by an accelerometer over the air ®lter case of theengine [7]. A cheap electret microphone along the exhaust pipe, conveniently housedto avoid direct contact with the exhaust gas, picked up the error signal [8]. A specialloudspeaker, located along the exhaust pipe in a side branch con®guration [9],withstanding operating temperatures of over 260�C, was used as secondary source.The adaptive ®lter to tailor the cancellation process was supplied by a program-mable, commercial ANC system.This ANC system successfully attenuated harmonics 2nd, 3rd, 4th, 7th, and 8th,

barely reduced harmonics 5th and 6th, but failed to cancel harmonics 9th, 10th, and11th. The ®rst harmonic was deliberately not cancelled to avoid instability problemswith the loudspeaker at this rather low frequency (about 30 Hz). The aim of thispaper is to elucidate the origin of this lack of attenuation at some harmonics and toalleviate it. Two possible causes are exhaustively analysed:

. The transfer function between the control source and the error sensor (thesecondary path).

. The coherence between the reference and error signals.

514 M. Cuesta, P. Cobo /Applied Acoustics 62 (2001) 513±526

Theoretical principles of active and passive control can be found elsewhere[2,4,5,10]. Here, the emphasis is on practical and technical aspects of the imple-mentation of a hybrid passive/active control system. Firstly, the experimental set-upis described. Then, the in¯uence of the exhaust system on the control performance isdiscussed. After that, the reference signal is optimised based upon its coherence withthe error signal.

2. Experimental set-up

A hybrid passive/active control prototype has been built to attenuate the noiseradiated by a Honda EG1900 generator. The passive system is composed of (Fig. 1):

Fig. 1. A view of the open enclosure showing the generator and the centrifugal fan.

M. Cuesta, P. Cobo /Applied Acoustics 62 (2001) 513±526 515

. The rectangular (100 � 71 � 53 cm) enclosure, built with 2 mm thick steelpanels, lined with 3 cm thick layer of absorbing material.

. The air intake opening, consisting on 8 mm diameter holes evenly distributedon a (29 � 18 cm) window.

. The gas exhaust system, consisting in two concentric tubes. Through the inner3 cm diameter steel tube ¯ows the hot gas. The outer 8 cm diameter sheet metaltube supports the ANC system. The inner tube penetrates 25 cm in the outertube. To avoid gas leaks the outer tube is sealed at the enclosure opening withrockwool. A more detailed description of the exhaust geometry is given inSection 3.

. Four passive mounts, consisting on three elastomeric layers of di�erent hardnessmatching each other in a sine pattern, inserted between the engine and theenclosure ¯oor. Each passive mount has dimensions (10 � 10 � 2.7 cm).

. A centrifugal fan inside the enclosure, in the opposite wall of the intake opening,to extract the hot air through a (13 � 10 cm) window. The fan runs at 1410rpm, supplies a volume ¯ow of 240 l/s, and its sound power level is 60 dBA.

. A temperature regulator, consisting of a temperature probe and an automaticcontroller, is installed for security reasons. If the temperature exceeds someprogrammed value, the engine is switched o�.

A single-input single-output (SISO) ANC system has been implemented on acommercial and con®gurable controller, following the recommendations of Oka-moto et al. [9]. The ANC system includes:

. The reference signal, picked up by one of two accelerometers on the air ®ltercase of the engine (Fig. 2). More details of the reference signal are given inSection 4.

. The control source, provided by a high temperature loudspeaker, in side-branch con®guration (Fig. 3). More details of the control source are given inSection 3.

. The error signal, picked up by one electret microphone on the junctionbetween the outer exhaust tube and the side-branch tube. To avoid directcontact with the exhaust gas the microphone is housed with a brass cap. ATe¯on ring isolates thermally the microphone from the brass cap.

Fig. 4 shows the ®nal layout of the hybrid control system.

3. In¯uence of the exhaust system

Engines usually generate high sound pressure levels in the exhaust system. In atypical SISO ANC system the secondary source (the loudspeaker in Fig. 4) musta�ord at the error microphone location an acoustic pressure equal (and opposite inphase) to that produced by the engine. If the amplitude of the signal yielded by theANC system and the power ampli®er is not enough to drive the loudspeaker to such


high pressure levels the ANC system will over¯ow. Furthermore, loudspeakersoperated at higher levels have shorter life and may distort. A combination of apassive mu�er with the ANC system can indeed lower the acoustic pressure, andhence, the power requirement of the secondary sources [11]. Okamoto et al. [9]demonstrated that a side-branch resonator with an end-mounted loudspeaker is thebest position for the secondary source because:

. An end-mounted high-impedance loudspeaker always requires less sourcestrength than a loudspeaker directly mounted at the duct wall.

Fig. 2. Location of accelerometers to pick up the reference signal.


Fig. 3. Two views of the enclosure showing the two exhaust systems.


. From a practical point of view, the loudspeaker can be better protected fromhostile (hot and dirty) environments.

A side-branch resonator will provide high transmission loss peaks at kL=(n+0.5)p, where k is the wavenumber and L is the side-branch length (quarter wavelength®lter). Thus, an active side-branch end-mounted controller will allow a large reduc-tion in the necessary secondary source strength around frequencies multiple of theresonance frequency.Fig. 5 shows the two exhaust systems designed for this research. In Case I, the

loudspeaker box connects directly with the 37 cm side-branch tube. In Case II, theloudspeaker connects to the 29 cm side-branch tube through a small Helmholtz

Fig. 4. Layout of the hybrid passive/active control system.

Fig. 5. Layout of the two tested exhaust systems (lengths in cm).


resonator. According to Okamoto et al. [9] a loudspeaker in front of a Helmholtzresonator driven at frequencies much higher than the Helmholtz resonance fre-quency is a close approach to a high-impedance source. Fig. 6 displays the exhaustlog-spectral noise level, measured at the error microphone, in both cases. The quarterwavelength frequency (f=c/4L) is around 230 Hz for Case I and 293 Hz for Case II.Therefore, the ®rst exhaust system provides higher transmission loss at harmonics4th to 9th. From Fig. 6 it is clear that the ®rst exhaust system provides higher passiveperformance.The ANC system must compensate for the transfer function between the loud-

speaker and the microphone. Thus, the simpler is this transfer function, the betterperforms the active ®lter. Fig. 7 shows the Bode plot of the secondary path in CasesI (dark line) and II (light line). It can be seen that the secondary source in the ®rstcon®guration supplies higher pressure levels at the error microphone in the wholefrequency band of interest (50±300 Hz). Thus, the ®rst exhaust system also yieldshigher active performance.In the following the in¯uence of the reference signal in the system performance is

analysed keeping just the ®rst exhaust system (Fig. 5a).

4. In¯uence of the reference signal

An adaptive ANC system generates at the error sensor location a replica of thereference signal, conveniently modi®ed by the transfer functions of the primary andsecondary paths. Therefore, the reference and the error signals should resemble oneanother. The resemblance between two signals in the frequency domain can bemeasured by the coherence function. The maximum reduction of the error spectrum,at frequency f, that would be achieved isÿ10 log10[1-Cre(f)], where Cre(f) is magnitude-squared coherence between the reference and the error signals [12]. Therefore,coherence is determinant for a high active control performance.In the absence of turbulence, a fully coherent reference signal should be provided

by a microphone along the exhaust duct. Unfortunately, the secondary loudspeakerwill generate sound propagating both upstream and downstream. Therefore, thecontrol source not only radiates antinoise which will interfere with the primary noisedownstream but also radiates upstream corrupting the reference signal (acousticfeedback). Feedback introduces poles in the ANC system that may instabilize theadaptive algorithm [12]. The problem becomes worse when the gain of the feedbackloop is too large, for instance in short ducts. When the primary noise is periodic, thisproblem can be avoided by using a synchronised nonacoustic sensor, such as anaccelerometer or tachometer.Fig. 8a shows the coherence of a single-reference signal picked up by an accel-

erometer on the air ®lter case of the engine (see Fig. 2) with the error signal, both®ltered from 40 to 300 Hz. Coherence seems to be acceptable for harmonics 2th, 3rd,4th and 10th, but likely will be insu�cient for the other harmonics. Fig. 9a displaysthe performance of the corresponding ANC system. A double-reference signal hasbeen synthesised by:


Fig. 6. Log-spectral level of the exhaust noise at the error microphones in Cases (a) I and (b) II.


Fig. 7. Bode plots of the secondary path transfer functions in Cases I (dark line) and II (light line).


Fig. 8. (a) Coherence between the only-reference signal and the error signal both ®ltered in 40±300 Hz; (b)

coherence between the double-reference signals ®ltered in 40±150 Hz (solid line) and 100±300 Hz (dashed

line) and the error signal ®ltered in 40±300 Hz.


Fig. 9. Log-spectra of exhaust noise at the error microphone with the ANC system o� (light line) and on

(dark line) when using (a) a single-reference and (b) a double-reference signal.


. Picking up a redundant vibration signal via two accelerometers located closeone another on the air ®lter case of the engine (see Fig. 2).

. Filtering both signals in overlapping frequency bands, namely, from 40 to 150Hz the ®rst, and from 100 to 300 Hz the second.

. Adjusting the gains of the preampli®ers so that both ®ltered signals haveapproximately equal amplitudes.

Fig. 8b shows the coherence of this double-reference signal with the error noise®ltered from 40 to 300 Hz. Higher coherence of all harmonics, except the 6th, arisesbetter ANC results (Fig. 9b). This demonstrates that the quality of the referencesignal is the main factor in¯uencing the performance of the designed hybrid system,and points out the path to introduce further improvements.

5. Summary and conclusions

A hybrid passive/active control system to attenuate the noise radiated by a smallgenerator, previously reported by the authors [1], has been optimised. The hybridsystem combines the performance of a passive enclosure to reduce the noise radiatedabove 400 Hz with the capability of an ANC system to cancel the periodic exhaustnoise below 400 Hz. The previous design successfully attenuated some harmonicsbut failed to reduce some others. Here, the origin of this performance lack has beenidenti®ed as the lost of coherence between the reference and the error signals at thefrequencies of these harmonics. To alleviate this problem a higher coherent referencesignal has been synthesised by combining the redundant information provided by twoaccelerometers closely located on the air ®lter case of the engine. The transfer functionof the secondary path, also analysed, has a minor in¯uence on the system performance.These results point out the direction to further improve the designed prototype.

Acknowledgements

This work has been supported by the CICYT, Project AMB99-1095-C02-01.

References

[1] Cuesta M, Cobo P. Active control of the exhaust noise radiated by an enclosed generator. Applied

Acoustics 2000;61(1):83±94.

[2] Beranek LL, VeÂ r IL. Noise and vibration control engineering. Principles and applications. New

York: John Wiley and Sons, 1992.

[3] Elliott SJ. Down with noise. IEEE Spectrum, 1999 June: 54±61.

[4] Nelson PA, Elliott SJ. Active control of sound. London: Academic Press, 1992.

[5] Hansen CH, Snyder SD. Active control of noise and vibration. London: E&FN Spon, 1997.

[6] Cobo P, Ranz C, Santiago JS, Pons J, Siguero M, Delgado C. Insertion loss measurements of an

acoustical enclosure by using sound power and MLS methods. J Acoust Soc Am 1998;103(5/2):3074.


[7] Cuesta M, Cobo P. Control activo de una fuente encapsulada utilizando una referencia muÂ ltiple.

Proceedings of Tecniacustica 1999, Avila, Spain.

[8] Cuesta M, Cobo P. A passive/active hybrid system for control of the noise radiated by a small gen-

erator. Proceedings of Forum Acusticum 1999, Berlin, Germany.

[9] Okamoto Y, Bodem H, Abom M. Active noise control via side-branch resonators. J Acoust Soc Am

1994;96(3):1533±8.

[10] Cobo P. Control Activo del Ruido. Principios y Aplicaciones. Madrid, CSIC, ColeccioÂ n Textos

Universitarios no. 26, 1997.

[11] Munjal ML, Eriksson LJ. Analysis of a hybrid noise control system for a duct. J Acoust Soc Am

1989;86(2):832±4.

[12] Kuo SM, Morgan DR. Active noise control systems. Algorithms and DSP implementations. New

York: John Wiley & Sons Inc, 1996.


Optimisation of an active control system to reduce the exhaust noise radiated by a small generator

Documents

Transcript of Optimisation of an active control system to reduce the exhaust noise radiated by a small generator