Source identiﬁcation and noise reduction of a ... · Source identiﬁcation and noise reduction of a reciprocating compressor; a case history N. Bert Roozen Philips Applied Technologies,

Source identification and noise reduction ofa reciprocating compressor; a case historyN. Bert RoozenPhilips Applied Technologies, High Tech Campus 7, 5656AE Eindhoven, The NetherlandsEindhoven University of Technology, P.O. Box 513, 5600MD Eindhoven, The Netherlands

Jozef van den OetelaarGEA Grasso, Parallelweg 27, 5223AL ’s Hertogenbosch, The Netherlands

Alex Geerlings

Theon VliegenthartPhilips Applied Technologies, High Tech Campus 7, 5656AE Eindhoven, The Netherlands

A noise source can be very complex in nature. In noise control engineering an essential first step is to identify thestrongest contributing noise sources. This paper discusses a practical case history, in which a series of measurementtechniques was applied to a reciprocating compressor to identify the strongest sound source. In addition numericalprediction methods were used to give direction towards a lower noise design of the compressor. Structural modi-fications to the compressor are discussed, leading to a significant reduction of the noise levels. The sound powerlevel of the reciprocating compressor was reduced by 5 dB. Moreover, from a perceptual point of view, customersexperience the sound as more robust and more pleasant as well, probably because of shifting frequencies.

1. INTRODUCTION

During the development of the Grasso V700 compressor a4 cylinder prototype was created consisting of bended steelplates welded together, as shown in Fig. 1. The plating for thecrankcase had a thickness of 15 mm, whilst the cylinder headswere constructed from 8 mm thick steel plate for the bendedparts and 15 mm plating for the end grains. The material usedwas steel S355J2+N.1 The new assembly is straightforward, al-lowing a more efficient production as compared to other com-pressor concepts. Besides, separated cylinder heads have ma-jor advantages in terms of thermodynamic behavior. However,it turned out that the steel plated compressor produced an un-wanted increase in noise of approximately 6 dB compared toits predecessor, the Grasso 412 compressor, which used a steelpipe welded cylinder head.

This paper describes a practical case history of the acousticanalysis and redesign of the compressor to restore former noiselevels. Design directions were based on in-situ measurementsand numerical simulations.

The paper is organized as follows. It starts in Section 2 withan experimental noise analysis, including sound pressure levelmeasurements and sound intensity measurements of the com-plete compressor to determine the most dominant frequencybands. This is followed by a partial noise source analysis bymeans of a surface intensity measurement technique to iden-tify the most dominant part of the machine contributing to thedominant frequency band. After that a modal analysis mea-surement and a modal analysis simulation are presented of thedominant part of the compressor. Finally, in Section 3 a struc-tural modification of the most dominant part of the compressoris discussed. Its implementation was acoustically validated bymeans of a sound intensity measurement of the complete com-

pressor, confirming a significant noise reduction. In Section 4the conclusions are given.

2. NOISE ANALYSIS

2.1. Compressor harmonics

For a compressor operating at 1200 rpm, the fundamentalfrequency of the noise spectrum corresponds to 1200/60=20Hz. A sound pressure level measurement was performed. Fig-ure 2 shows that the compressor noise was dominated by thefundamental rotational frequency and its harmonics. Only thefundamental and (many) harmonic frequencies of the rotationspeed contribute to the spectrum, covering the whole audiblefrequency range. Therefore, shifting operational speed in orderto avoid structural resonances is obviously not an option. Gen-erally speaking, excitation forces are rarely purely sinusoidalin form, so that harmonics appear naturally as can be explainedby Fourier series expansion. In this specific case of a compres-sor it is caused by the fact that the compression in the cylinderis far from smooth, causing an abrupt increase of the pressureinside the cylinder. Moreover, valve noise also attributes tothe noise spectrum, which contains many harmonics becauseof its impulse repetitive character. Excitation due to discreteharmonics is typical in many structural acoustic cases.

Though the acoustic noise consists only of discrete frequen-cies, the dynamics of the structural components of the com-pressor play an important role. Resonant behavior of structuralcomponents of the compressor may increase the structural re-sponse. However, the presence of structural resonances is nowmore obscured compared to cases in which a mechanical sys-tem is excited by a more broadband spectrum.

90 (pp. 90–98) International Journal of Acoustics and Vibration, Vol. 14, No. 2, 2009

N. Bert Roozen, et al.: SOURCE IDENTIFICATION AND NOISE REDUCTION OF A RECIPROCATING COMPRESSOR; A CASE HISTORY

Figure 1. Prototype Grasso V700.

0 500 1000 1500 200020

30

40

50

60

70

80

Frequency [Hz] ∆f=1.56Hz

Un

wei

gh

ted

So

un

d P

ress

ure

LP [

dB

] re

20

µPa

Figure 2. Typical sound pressure level at a distance of 1 m, compressor oper-ating at 1200 rpm.

2.2. Total Sound powerSound intensity measurements of the complete compressor

were performed to identify the most dominant frequency band.We measured the overall noise spectrum and ranked the var-ious A-weighted frequency octave band contributions. Thesound intensity measurements were performed according toISO 9614-2 (intensity scanning)2 on the Grasso 412 and proto-type Grasso V700, using a Bruel & Kjaer sound intensity probekit type 3581, intensity microphone pair type 4181 and a Bruel& Kjaer Pulse data acquisition system. The repeatability checkand the sound field pressure-intensity indicator check, indi-cating the quality of the sound intensity measurement, wereperformed, see Section 3.3. The radiated sound power is pre-sented in Fig. 3, comparing the Grasso V700 prototype withits predecessor, the Grasso 412 compressor, showing the un-wanted increase of A-weighted noise levels of approximately6 dB. For both the Grasso V700 prototype and the Grasso 412compressor the strongest contributions were found in the 1kHzoctave band.

2.3. Partial Noise Source AnalysisThe next step is a combined operational deflection shape

analysis3, 4 and a surface intensity analysis.5 Surface intensity

Figure 3. A-weighted octave Soundpower Spectrum of Grasso 412 (baseline)and prototype Grasso V700 (proto).

analysis is a promising technique for sound power measure-ments of individual noise sources in the presence of a noisysound field,6 which was certainly true in the present case dueto the complex sound radiation from parts of the compressor.Furthermore, Thrane et al. argue that the combination of bothtypes of measurement techniques can be very useful to performefficient noise control.7 In this paper we follow that approachby using the surface intensity analysis technique to identifythe strongest noise contributors, and the operational deflectionshape analysis to judge where and how to decrease the vibra-tion levels and associated noise radiation.

Figure 4 shows the test-set up in which the operational de-flection shape of the cylinder head was measured. For the op-erational deflection mode analysis we used one reference ac-celerometer and one roving accelerometer from Bruel & Kjaer,type 4383 combined with a Bruel & Kjaer Nexus conditioningamplifier type 2691 (two charge channels, two acoustic chan-nels). In total 30 measurement positions were defined on thecylinder head, with a point to point distance of 12 cm in ax-ial direction and 8 cm in circumferential direction, see Fig. 4.The roving accelerometer had a magnetic foot and could eas-ily be placed elsewhere. Since the cylinder heads were quitewarm a glove had to be used to move the roving magnetic ac-celerometer around. The allowable temperature range of the4383 accelerometer is from -74 oC up to 250 oC, which wassufficient for measurements on the cylinder heads which typi-cally reach a maximum temperature of 180 oC. Mass loadingeffects of the accelerometer were negligible for the 8 to 15 mmsteel plating, as will be discussed in Section 2.4.

2.3.1. Phase shift errors

The phase between the roving accelerometer and the refer-ence accelerometer was stored such that the phase informationover the studied partial surface was retained. In this case it wasrelatively simple to add also a roving microphone and store thephase information of the acoustic field as well. The micro-phone, a Bruel & Kjaer type 4192 combined with a preampli-fier type 2669 and a Nexus conditioning amplifier type 2691(two charge channels, two acoustic channels), was positionednear the accelerometer by means of a microphone stand, at adistance of 2.5 cm to the plate structure. This is significantlyless than one quarter of the acoustic wavelength for the fre-quencies of interest. Combining the surface velocities basedupon the accelerometer data with the measured surface pres-

International Journal of Acoustics and Vibration, Vol. 14, No. 2, 2009 91


Figure 4. Surface intensity measurement set-up on the compressor cylinderhead.

sures an estimate can be obtained of the local sound intensitynear the surface. This technique is referred to as the surfaceacoustical intensity method.5 An estimate of the partial radi-ated soundpower of the individual plate segments was obtainedby integrating the acoustic intensity over the surface.

A source of errors in the surface acoustical intensity methodis the phase errors, which can be caused by the instrumentationand by the spacing between the noise radiating structure andthe microphone, thus introducing an acoustic time lag. Fol-lowing McGary and Crocker the acoustic intensity vector ~Ican be determined from the one-sided cross-spectral density ofpressure and acceleration, Gpa, as follows, taking these phaseerrors into account:5∣∣∣~I∣∣∣ =

12π

∫ ∞0

Im(Gpa(f))cos(φ) +Re(Gpa(f))sin(φ)f

df

(1)where φ is the phase shift error and f denotes frequency.Re and Im denote the real and imaginary parts of a com-plex quantity, respectively. If there is no phase shift error,φ = 0, the acoustic intensity is proportional to the imagi-nary part of Gpa: I ∝ Im(Gpa)/f . If a significant phaseshift error is present, the true intensity is proportional toI ∝ {Im(Gpa)cos(θ) +Re(Gpa)sin(θ)} /f . Thus, follow-ing McGary and Crocker, it is convenient to define the error inintensity in decibels by the following error equation5

ERROR = 10log {cos(φ)+Re(Gpa)/Im(Gpa)sin(φ)} (dB) (2)

As concluded in the work of McGary and Crocker the error inthe intensity depends not only on the phase shift φ itself, butalso on the ratio of the real to the imaginary part of the crossspectrum Gpa.

Using the measured cross spectra Gpa, spatially averagedover the 30 different measurement positions, Eq. (2) can beused to investigate the effect of phase errors on the sound in-tensity estimate, for this specific situation. The error in theintensity caused by an uncorrected phase shift is shown in Fig.6, which indicates for this specific situation that the impact ofthe phase errors is not very large for phase shifts between -30o

and +90o. This is due to the fact that the reactive part of thesound intensity is relatively small as compared to the active

(a)

(b)

Figure 5. Experimental results under operational conditions, a) operationaldeflection shape at 1020 Hz, b) operational pressure distribution at 1020 Hz.

part for this specific case, meaning that the real part of Gpa ismuch smaller than the imaginary part of Gpa. The fact that thereactive part appears to be small logical as in Section 2.3.2 itwill be concluded that the structural wavelength is larger thanthe acoustic wavelength. This is typical for an efficient soundradiator.

According to the technical documentation of the measure-ment equipment, the phase characteristics of the Bruel & KjaerNexus 2691 conditioning amplifier (±0.1o), 4383 accelerom-eter (±2o) and 2669 microphone preamplifier (±0.1o) are rel-atively small within the frequency range of interest. The mi-crophone can cause a more significant phase error. The typi-cal phase characteristic of the Bruel & Kjaer microphone type4192 is shown in Fig. 7, as specified by the manufacturer.8

From this figure it can be seen that for frequencies between100 and 2000 Hz the phase error is less than 6o. Thus, it canbe concluded that the phase errors due to the instrumentationare relatively small within the frequency range of interest ascompared to the phase shift caused by the acoustic time lagthat is introduced by the distance of 2.5 cm between the mi-crophone and radiating structure, which would be about 30o at

92 International Journal of Acoustics and Vibration, Vol. 14, No. 2, 2009


−90 −60 −30 0 30 60 90−25

−20

−15

−10

−5

0

5

Phase shift φ (degrees)

Err

or in

inte

nsity

(dB

)

Figure 6. Error in intensity caused by uncorrected phase shift.

Figure 7. Typical phase response of a Bruel & Kjaer microphone type 4192.8

1000 Hz.Fortunately, for this specific situation we have a high active

and a low reactive sound field, in which case, according to Fig.6, the impact of the phase errors between 0o and +60o resultsin an error of less than 1 dB. For that reason it is sufficient totake only the acoustic time lag into account, using a phase shiftof

φ = 2π∆l/λ = 2π∆lf/c (3)

where ∆l is the distance between the microphone and the radi-ating surface onto which the accelerometer is mounted, givinga phase shift of approximately +30o at 1000 Hz for ∆l = 2.5cm. However, in general situations with both a high active andreactive sound field, one should take into account the phaseshift due to both the instrumentation and the acoustic time lagas the impact of small changes in the phase shifts on the accu-racy of the surface intensity results can be much larger in thosecases.

2.3.2. Results

The operational deflection shape of the cylinder head struc-ture and the pressure distribution across the cylinder head atthe frequencies near 1000 Hz are shown in Fig. 5. The struc-tural deflection shape shows only a few half wavelengths in thecylinder head circumference, and only a single half wavelengthin the longitudinal direction of the cylinder head. From Fig. 5the structural wavelength near 1kHz is estimated to be approx-imately 55 cm. Compared to the acoustic wavelength in air atthis frequency, λ = c/f =34 cm, the structural wavelength islarger.

Joining these measurements in proper phase relation, tak-ing into account possible phase shifts, an estimate can be ob-

Proto

Cylinderhead (estimate)

Lw [A] = 102.3 dB re 1pW

Lw [A] = 102.6 dB re 1pW

16 31.5 63 125 250 500 1k 2k 4k 8k 50

60

70

80

90

100

110

Octaveband Centre Frequency [Hz]

A−w

eigh

ted

Sou

ndpo

wer

[dB

]

Figure 8. A-weighted partial sound power estimate of the cylinderhead com-pared to the overall A-weighted octave sound power spectrum of the GrassoV700 compressor (proto).

tained of the active sound intensity close to the radiating cylin-der head using Eq. (1).

Eq. (3) is used for the phase error between the microphoneand accelerometer. This yields sound intensity estimates inradial direction at the 30 measurement positions. Integratingthe sound intensity over the cylinder head surface gives a par-tial sound power estimate. The partial sound power estimatefor the two cylinder heads is shown in Fig. 8. Frequenciesabove 2000 Hz were not available from the operational deflec-tion shape measurement data. Fortunately, above 2000 Hz thetotal sound power of the compressor is not significant.

From Fig. 8 it can be seen that at the 1kHz acousticallydominant frequency band, the cylinder head contributes sig-nificantly. Theoretically, the partial sound power contributionscannot exceed the total sound power of a machine. Appar-ently, this is violated at 1kHz by 2 dB (with a 0.3 dB total leveldifference), which is probably due to measurement errors. Inconclusion, it can be stated that the cylinder heads dominate.

2.4. Driving Point Measurements

To obtain information on possible resonant behavior of thesound radiating cylinder head, a driving point measurementwas performed. By means of a hammer the structure is ex-cited, and the structural response is measured by means of anaccelerometer, as illustrated in Fig. 9. This time the machine isnot in operation and therefore the cylinder heads have an am-bient temperature. The ratio of the structural velocity v and theforce F exerted by the hammer is calculated and presented asthe structural mobility

Y = v/F (4)

A high mobility denotes a ”flexible” structure, whereas a lowmobility denotes a ”stiff” structure. A high mobility Y mayindicate contributing radiators.

The driving point measurements were performed using ahammer, which was instrumented with an Aptech force trans-ducer, type AU-01. A Bruel & Kjaer accelerometer type 4383was placed nearby the point of excitation to measure the struc-tural response due to the hammer excitation. Both signals werefed into a Bruel & Kjaer Nexus conditioning amplifier type2691.

The effects of mass loading of the accelerometer will nowbe considered. The classical mass cancelation procedure for



Figure 9. Cylinder head driving point measurement set-up.

200 400 600 800 1000 1200 1400 1600 1800 200010

−6

10−5

10−4

10−3

Frequency [Hz] ∆f=1.56Hz

ms

−1N

−1

discharge connectioncylinder headcrank casecylinder head sidewallseparator

Figure 10. Driving point mobility measurements taken at center of a numberof plate-like structural parts of the compressor.

the driving point accelerance is given by9, 10

1H

=1

Hmeas−m (5)

where H is the true accelerance of the structure, Hmeas is themeasured accelerance and m is the mass of the accelerometer.In terms of point mobility Eq. (5) can be written as

1Y

=1

Ymeas− iωm (6)

where ω is the angular frequency, ω = 2πf and i =√−1.

Defining the error due to the mass loading of the accelerometeras (1− Ymeas/Y ) · 100%, we can write∣∣∣∣1− Ymeas

Y

∣∣∣∣ · 100% = ωm|Ymeas| · 100% (7)

Knowing that the mass of the accelerometer, a Bruel & Kjaertype 4383, is less than 20 grams, and looking at the measuredpoint mobilities as shown in Fig. 10 it can be calculated thatup to a frequency of 1400 Hz the error due to mass loading isless than 2%.

Figure 10 shows the measured structural mobility of a num-ber of plate-like parts of the compressor, including the cylinder

Figure 11. Cylinder head modal analysis measurement set-up, showing the 30measurement positions.

heads, the crankcase and the separator (see Fig. 1). Knowingthat the compressor is acoustically most dominant in the 250up to 1000 Hz 1/3rd octave frequency bands, as shown in Fig.3, it is likely that the relatively high mobility of the cylinderheads in the frequency range from 200 – 1200 Hz mark them asacoustically dominant contributers. This observation is also inagreement with the finding of the partial sound power estimatemeasurements as discussed in the previous section. However,it is stressed that mobility in itself is not sufficient to identifydominant sources. It adds to the insight and gives an indica-tion, knowing the pressure field inside the machine is more orless uniformly distributed.

2.5. Modal Analyses (Measurements)

Having identified the cylinder heads as the dominant noisesource we could support a design modification of the cylinderheads by a detailed modal analysis11 of the cylinder head. Theearlier defined grid of 30 points, with a point to point distanceof 12 cm in axial direction and 8 cm in circumferential direc-tion, was sufficient to capture the first relevant modes (see Fig.11). One accelerometer was fixed and the other accelerome-ter was roving across the 30 measurement positions. Thus, 30transfer functions from the reference accelerometer to the rov-ing accelerometer type 4383 were obtained, from which thenormal modes of the structure were extracted.

Figure 12 shows two structural mode shapes having reso-nance frequencies near 1070 Hz and 1300 Hz. The 1070 Hzmode is most likely contributing significantly to the soundpressure level. The 1070 Hz mode has one half wavelength inthe axial direction and three half wavelengths in circumferen-tial direction. The 1300 Hz mode has also one half wavelengthin axial direction but four half wavelengths in circumferentialdirection.



(a)

(b)

Figure 12. Experimental modal analysis results, Grasso V700 prototype a)natural mode at 1070 Hz, b) natural mode at 1300 Hz.

2.6. Modal Analysis (Simulations)The structural dynamics of the cylinder heads were simu-

lated with the finite element method,12,13 using the softwarepackage Pro-Mechanica 3.0. The model was able to predictmost of the experimentally obtained normal modes, of whichtwo modes are shown in Fig. 13. These two normal modes cor-respond well to the experimental modes shown in Fig. 12, thusvalidating the finite element model, including material proper-ties and boundary conditions.

3. NOISE REDUCTION

Noise control of machinery is well described in literature.13

In this specific case the main focus was to get rid of the

(a)

(b)

Figure 13. Predicted natural modes, Grasso V700 prototype, a) natural modeat 1081 Hz, b) natural mode at 1285 Hz.

efficient natural modes of the steel plated cylinder headat approximately 1000 Hz and decrease the flexibility ofthe remaining modes. Alternatively, the noise path couldbe damped by means of double plating. It was decided toabandon the idea of double plating because of possible majorproblems with weldability and complying with the PressureEquipment Directive (97/23/EC).14

Amongst many others, the following design changes werediscussed:

• Thickness increase (e.g. from 8 mm to 15 mm) of thehead would result in a stiffer (increase of factor 6.6) andheavier (almost doubled) construction. As a result the nat-ural frequencies could be shifted upwards roughly by afactor 2. The added weight reduces the deflection levelsdue to the excitation. However it was questioned if thiswould be enough to counteract any increase in radiationefficiency. More practical, the cylinder head would be tooheavy to handle.

• Use of stiffening ribs on the outside of the cylinder head(for better weldability) would also raise the natural fre-quencies, but then the radiating surface would increasewhile the mass effect would be lower. Also in this case apotential danger for an increased radiation efficiency ex-ists.



Figure 14. Base shape of cylinder head. Left: old design (compare Fig. 11).Right: new design (compare Fig. 15).

Figure 15. Ductile cast iron cylinder head driving point measurement set-up.

• Changing the production method towards casting, sinceductile cast iron also adds more internal damping(roughly factor 4). Appropriate material in respect to thePressure Equipment Directive14 would be ductile cast iron(EN-GJS-40015), taken into account the strength/safetyfactor ratio. However, one should be aware that highersafety factors together with lower mechanical propertiescompared to steel (S355J2+N1) would probably resultin higher mass and possibly additional mounting covers.Nevertheless, much more constructive possibilities couldbe explored in order to cope with requirements regardingsound, flow efficiency, heat transfer, etcetera. By chang-ing the base shapes when using cast iron from a half-pipeto a coupled pipe (see Fig. 14), the natural frequency ofthe efficiently radiating modeshape was shifted to higherfrequencies. Simultaneously the flexibility was decreasedsignificantly.

To reduce the noise of the steel plated cylinder head it wasdecided to redesign the head using ductile cast iron. Ductilecast iron has inherently much more material damping, whichshould damp the resonant behavior, which was one of the mainproblems in the prototype. Furthermore, it was decided to con-struct the cylinder head more compact, thus decreasing the mo-bility of the head structure.

Supported by modal analysis simulations by means of finiteelements, the cylinder heads were redesigned. In the next sec-tion the results of the finite element simulations are presented.

(a)

(b)

Figure 16. Numerical analysis cast iron cylinder head a) solid model, b) nor-mal mode at 1284 Hz.

3.1. Finite Element Analysis RedesignedCylinder Heads

Due to the more compact design of the cylinder heads, asshown in Fig. 16a the resonance frequencies were shifted up-wards. The finite element model predicted the mode as shownin Fig. 13a to be shifted up from 1000 Hz to 1320 Hz, asshown in Fig. 16b. As the excitation spectrum level decreaseswith increasing frequency, this will reduce the acoustic noiseproduction.

3.2. Verification Driving Point Measure-ments On Improved Prototype

To verify the decrease in mobility for the redesign of thecompressor head, driving point measurements were performedon the Grasso V700 pre-production compressor. Figure 15shows the test set-up of the driving point measurements on thecast iron cylinder head. The results of these measurements areshown in Fig. 17. As compared to the prototype cylinder head,shown with a dotted line, the mobility was reduced by almosta factor 10 (20 dB) at frequencies around 1000 Hz. Moreover,for frequencies above 800 Hz the dynamic behavior was muchmore damped, though between 400 and 700 Hz there seems tobe some resonant behavior present.



200 400 600 800 1000 1200 1400 1600 1800 200010

−6

10−5

10−4

10−3

Frequency [Hz]

ms

−1N

−1

Cast iron cylinder headSteel plated cylinder head

Figure 17. Driving point mobility prototype and redesigned cylinder head.

Figure 18. Compressor sound power level of Grasso 412 (baseline), prototypeGrasso V700 (proto) and pre-production Grasso V700 (final).

3.3. Validation Intensity Measurements (ISO9614)

To quantify the achieved noise reduction, sound intensitymeasurements according to ISO 9614-2 (intensity scanning)2

were again performed on the new prototype (see Fig. 19). ABruel & Kjaer sound intensity probe kit type 3581, intensitymicrophone pair type 4181 and a Bruel & Kjaer Pulse data ac-quisition system were used. The sound power of the sourcewas determined by taking the surface integral across the mea-surement surface of the normal component of the sound inten-sity vector.

The quality of a sound intensity measurements is indicatedby the repeatability index and by the sound field pressure-intensity index,2,16.17 The repeatability check consists of twosubsequent measurements of the sound power through a mea-surement surface, scanning with the sound intensity probe inperpendicular oriented scan patterns. The check requires thedifference between the two power estimates to be smaller than3 dB for each 1/3rd octave band and for each scanned sub area.For the 1/3rd octave bands above 100 Hz this requirement wassatisfied in almost all individual sub area measurements. Forthe 1/3rd octave bands below 100 Hz, however, it appeared tobe very difficult to comply. Fortunately, in absolute sense thesound power levels in the 50 Hz, 63 Hz and 80 Hz 1/3rd oc-tave bands were not high, and therefor not significant in termsof the overall sound power level of the compressor.

(a)

(b)

Figure 19. Sound intensity measurement set-up (Grasso V700 pre-productionwith cast iron cylinder heads).

The sound field pressure-intensity indicator checks for theerror due to phase mismatches in the measurement chain. Theratio between the sound pressure level and the sound inten-sity level is called the pressure-intensity index. Two types ofpressure-intensity indices exists, i.e. the pressure-intensity in-dex of a measurement, called the sound field pressure-intensityindex LpI , and the pressure-intensity index as measured dur-ing a calibrate test, the residual pressure-intensity index LpI0.To obtain an accuracy better than 1 dB due to phase mismatch,a minimum difference of 7 dB between the residual pressure-intensity index LpI0 and the sound field pressure-intensity in-dex LpI is required (Grade 3). A minimum difference of 10dB ensures a maximum error of 0.5 dB (Grade 2).

To improve the quality of the sound intensity measurements,i.e. to reduce the sound field pressure-intensity index LpI ,the reverberation of the test cell was reduced by means of 10cm thick foam walls during the verification measurements, asshown in Fig. 19). Nevertheless, it appeared that Grade 3was the best grade that could be obtained for frequencies atand above 250 Hz, probably due to a relatively low residualpressure-intensity index LpI0. Thus an accuracy of 1 dB interms of sound power could be obtained, which might be alsoone of the reasons for the discrepancy found in Fig. 8.

The measured sound power radiated by pre-production (withthe redesigned cast iron cylinder head) as well as the prototype(steel plated cylinder head) and its predecessor, the Grasso 12



compressor, are presented in Fig. 18. The 1000 Hz 1/1 octaveband sound power levels for the cast-iron pre-production com-pressor were reduced by 7 dB, resulting in an overall reductionof the A-weighted sound power level in the order of 5 dB ascompared to the steel plated cylinder head V700 prototype. Ascompared to the predecessor, the Grasso 12 compressor, theGrasso V700 pre-production emits an insignificant 1 dB moresound power. From Fig. 18 it can also be seen that the castiron cylinder head V700 pre-production compressor produceshigher noise levels at frequencies below 500 Hz, which, from aperceptual point of view, customers experience as more robustand more pleasant.

4. CONCLUSIONS

1. The combination of signal analysis (excitation harmonics,operational deflection shape, surface intensity) and sys-tem analysis (modal analysis) is an essential approach fora first-time-right analysis. The surface intensity measure-ment contributed to this by clearly pinpointing the domi-nant noise source.

2. The unwanted noise increase of approximately 6 dB dueto a new plated cylinder head design was successfully re-solved. By means of a proper use of noise control en-gineering insights and analysis tools the dominant noisesource in the cylinder head was identified. Using thisknowledge the cylinder head was completely redesignedin cast iron which led to a 5 dB noise reduction.

3. From a perceptual point of view, probably because ofshifting frequencies, customers characterize the sound it-self as more robust and more pleasant as well.

REFERENCES1 DIN EN 10025-2, Hot rolled products of structural steels

- Part 2: Technical delivery conditions for non-alloystructural steels, DIN-adopted European Standard, (2005)(2004).

2 ISO 9614-2, Acoustics - determination of sound power lev-els of noise sources using sound intensity - Part 2: Mea-surement by scanning. International Organization for Stan-dardization, (1996).

3 Dossing, O., Structural Stroboscopy - Measurement of Op-erational Deflection Shapes, Sound & Vibration Magazine,(1988).

4 Richardson, M. H., Is it a mode shape, or an operating de-flection shape?, Sound & Vibration Magazine, (1997).

5 McGary, M. C. and Crocker, M. J., Phase shift errors inthe theory and practice of surface intensity measurements,Journal of Sound and Vibration, 82(2), 275-288, (1982).

6 Yoshihiro Hirao, Kohei Yamamoto, Kentaro Nakamura andSadayuki Ueha, Development of a hand-held sensor probefor detection of sound components radiated from a spe-cific device using surface intensity measurements, AppliedAcoustics, 65(7), 719-735, (2004).

7 Thrane, N. and Gade, S., Use of operational deflectionshapes for noise control of discrete tones, Bruel & Kjaer,Technical Review, (1988).

8 Microphone Handbook for the FalconTM Range of Micro-phone Products, Bruel & Kjær BA510512 (1995).

9 Cakar, O. and Sanliturk, K. Y., Elimination of transducermass loading effects from frequency response functions,Mechanical Systems and Signal Processing, 19(1), 87-104,(2005).

10 Silva, J. M. M., Maia, N. M. M. and Ribeiro, A. M. R., Can-cellation of mass-loading effects of transducers and evalu-ation of unmeasured frequency response functions, Journalof Sound and Vibration 236(5), 761-779, (2000).

11 D.J. Ewins, Modal testing: theory, practice and application(2nd edtion), Research Studies Press Ltd., Great-Britain,(2000).

12 K.J. Bathe and E.L. Wilson, Numerical methods in finite el-ement analysis, Prentice-Hall Inc., Englewood Cliffs, NewJersey, (1976).

13 Crocker, M. J. (Ed.), Handbook of noise and vibration con-trol, John Wiley & Sons, Inc. (2007).

14 97/23/EC, The European Pressure Equipment Directive,(1997).

15 DIN EN 1563, Founding - Spheroidal graphite cast irons,(2008) (2005).

16 Crocker, M. J. and Arenas, J. P., Fundamentals of the Di-rect Measurement of Sound Intensity and Practical Appli-cations, Acoustical Physics, 49(2), 163-175, (2003).

17 Gade, S., Validity of Inensity Measurements in PartiallyDiffuse Sound Field, Bruel & Kjaer Technical Review, No.4. (1985).


Documents

Source identiﬁcation and noise reduction of a ... · Source identiﬁcation and noise reduction of a reciprocating compressor; a case history N. Bert Roozen Philips Applied Technologies,