You are on page 1of 8

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/11036137

A jet engine noise measurement and prediction


tool

Article in The Journal of the Acoustical Society of America December 2002


DOI: 10.1121/1.1506690 Source: PubMed

CITATIONS READS

10 206

5 authors, including:

Abdelkader Frendi Ten-See Wang


University of Alabama in Huntsville NASA - Marshall Space Flight Center
89 PUBLICATIONS 469 CITATIONS 118 PUBLICATIONS 928 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Propulsion View project

Fundamentals of combustion View project

All content following this page was uploaded by Abdelkader Frendi on 13 July 2017.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
A jet engine noise measurement and prediction tool
Abdelkader Frendia)
Mechanical and Aerospace Engineering Department, University of Alabama in Huntsville, Huntsville,
Alabama 35899

Wade D. Dorland and Thein Maung


AI Signal Research, Inc., Huntsville, Alabama 35805
Tom Nesman and Ten-See Wang
NASA Marshall Space Flight Center, Huntsville, Alabama 35812
Received 8 October 2001; revised 14 June 2002; accepted 22 July 2002
In this paper, the authors describe an innovative jet engine noise measurement and prediction tool.
The tool measures sound-pressure levels and frequency spectra in the far field. In addition, the tool
provides predicted results while the measurements are being made. The predictions are based on an
existing computational fluid dynamics database coupled to an empirical acoustic radiation model
based on the far-field approximation to the Lighthill acoustic analogy. Preliminary tests of this
acoustic measurement and prediction tool produced very encouraging results. 2002 Acoustical
Society of America. DOI: 10.1121/1.1506690
PACS numbers: 43.58.Kr, 43.58.Ta SLE

I. INTRODUCTION cant contributions to jet noise theory have been made by


various researchers and are summarized in the review papers
In the last 20 years, tremendous progress has been made by Ffowcs Williams,7,8 Goldstein,9 and Tam.10
in computational fluid dynamics CFD. This progress could Mani et al.11 used Ribner4 and Ffowcs Williams12 jet
not have been achieved without the quantum leap in com- noise models in their extensive studies on high-velocity jet
puter power. In addition, computational algorithms have be- noise. Good agreement was obtained between the model and
come very efficient and robust. However, with all these ad- the experimental results for various nozzle geometries and
vancements, CFD has remained a tool used by engineers and flow regimes. More recently, Khavaran et al.13 used a (k
scientists mainly to understand experimental observations. In ) turbulence model to compute the jet mixing noise from
recent years, however, there have been few cases where CFD a supersonic axisymmetric convergingdiverging nozzle.
was used extensively in the research and development They used Mani et al.s11 approach to arrive at their results.
phase.1 This gave CFD a vote of confidence from the engi- Comparisons with experiments showed good agreement.
neering community, which in itself is a great achievement.
Other computational studies using the (k) turbulence
However, using CFD for real-time predictions has not been
model were carried out for example, Refs. 14 and 15 with
attempted because of its slow turnaround time. One way to
varying degrees of success.
overcome this delay is to create a large CFD database with
Recently, Tam et al.16 proposed a self-contained, semi-
various parameters.
empirical jet noise theory for the prediction of fine-scale tur-
In this paper, we present a detailed description of a mea-
surement and prediction tool that can be used during a real- bulence noise from high-speed jets. This theory also uses jet
time experiment. This tool uses a CFD database, made up of flow turbulence information supplied by a (k) turbulence
results from various Reynolds-averaged-NavierStokes model. In addition to the empirical constants found in the
computations, a semiempirical far-field acoustic model, and turbulence model, Tam et al.s jet noise model introduces
the experimental results being measured. Results, both mea- three additional empirical constants that need to be fixed
sured and predicted, are displayed on the same screen in real using experimental data. The resulting noise predictions ex-
time. The semiempirical far-field acoustic model2 is derived hibited excellent agreement with measurements. This theo-
from the Lighthill acoustic analogy.3 This equation gives the retical approach uses linearized Euler equations to determine
sound pressure at a far-field point as radiated by a localized the far-field acoustic pressure, which is a departure from the
unsteady source or turbulent flow. Ribner4 used the Lighthill widely used Lighthill acoustic analogy. However, Tam
acoustic analogy to arrive at a relation between the mean- et al.s model cannot be used for real-time predictions be-
square pressure radiated by a jet plume into the far field as cause of its computational cost and time requirements.
the integral over the plume volume of quadrupole correla- In the remainder of the paper, the authors describe the
tions. Ribner5,6 also showed the importance of convection semiempirical acoustic model used, CFD database results,
and refraction on the far-field jet noise. Several other signifi- experimental measurements, and, finally, the acoustic model
validation. We close the paper with a description of the inte-
a
Author to whom correspondence should be addressed. Electronic mail: grated measurement and prediction tool and related conclu-
frendi@mae.uah.edu sions.

2036 J. Acoust. Soc. Am. 112 (5), Pt. 1, Nov. 2002 0001-4966/2002/112(5)/2036/7/$19.00 2002 Acoustical Society of America
rR
C . 5
R
A convenient transformation of Eq. 3 is see Ribner4

p R, ,
2
R iR jR kR l
16 2 C 4 R 6
4
V
4 v i v j v k v l d r.
3

6
When the turbulence is assumed quasi-incompressible, Eq.
6 becomes
2R iR jR kR l

4
p 2 R, , v i v j v k v l d 3 r, 7
16 2 C 4 R 6 V
4

where S i jkl v i v j v k v l is the fourth-order velocity correla-


tion tensor. Ribner4 expressed S i jkl in terms of linear combi-
nations of second-order correlations as follows:
S i jkl S ik S jl S il S jk S i j S kl , 8
where S i j v i v j . Ribner has shown that for an axisymmet-
4

ric source field and for a locally isotropic turbulent field,


FIG. 1. Reference coordinate system.
only few of the 81 possible combinations of indices (i, j,k,l)
have a significant contribution. The various source terms are
then evaluated using the isotropic turbulence model of
II. THE SEMIEMPIRICAL MODEL Batchelor17 and by assuming that the second-order correla-
tions are separable functions of r and see Ribner4
To formulate the mathematical model, we begin with the
far-field approximation of the Lighthill acoustic analogy,3 S i j r, i j r g , 9
which can be written as with


2
R iR j Tij 3 i j r k f 21 r f i j 21 f r i r j /r , 10
p R,t d y, 1
4 C 2 R 3 V t2 where k is the local turbulent kinetic energy and f some uni-
where T i j is the Lighthill stress tensor given by versal function of r given by
2 r 2 /L 2
T i j v i v j i j C 2 i j . 2 f r e , with r 2 x 2 y 2 z 2 . 11
In the above definition of f, L is a longitudinal macroscale of
In Eq. 2, the dominant term is the unsteady momentum flux
the turbulence. In Eq. 9, g( ) is the time delay factor as-
v i v j , i j is the viscous compressive stress tensor, the
sumed by Ribner4 to be of the form
local pressure, the density, C the ambient, adiabatic speed
of sound, v i the velocity, and i j is the Kronecker symbol 2
g e / 0 , 12
0 if i j, 1 if i j). The notation in Eq. 1 de-
notes evaluation at the retarded time, tR/C . On re- where 0 represents the characteristic time delay which de-
taining the dominant term, i.e., v i v j , the mean-square pres- termines the minimum significant correlation in a moving
reference frame. The length scale L in Eq. 11 is related to
sure p 2 radiated at an observer point R,, in polar
coordinates see Fig. 1 can be written as
0 by
Lc l k 1/2 0 , 13
p 2 R, ,
where c l is an empirical constant. Using Eqs. 813 in Eq.


RiR jRkRl
16 2 C 4 R 6
V V
2 v i v j 2 v k v l 3
t 2 t 2
d y d 3 y . 3
7 and taking the Fourier transform of the resulting equation
with respect to , we arrive at the spectrum
2
The first term under the overbar is evaluated at (y ,t ) while I c 3l 2 k 7/2 0 4 e 1/8 0 , 14
the second term is evaluated at (y ,t ). The emission times where is the source frequency, which is related to the
(t ,t ) are retarded with respect to the reception time t. The observed frequency by see Mani et al.11
quadrupole correlation shown with an overbar can be ex-
pressed as a function of the midpoint y and the separation in 2 f 1M c cos 2 c d k 1/2/C 2 . 15
space and time In Eq. 15, M c is the convection Mach number given by
y y y ,
1
2 ry y , t t . 4 M c 0.5M c c M j , 16
If the observer distance R is large compared to the flow where M is the local Mach number, M j is the nozzle exit
dimensions, then Mach number, and (c c ,c d ) are empirical constants. Light-

J. Acoust. Soc. Am., Vol. 112, No. 5, Pt. 1, Nov. 2002 Frendi et al.: Jet engine noise measurement and prediction 2037
hills acoustic analogy approach, which is based on the clas-
sical wave equation, doesnt include the effects of the sur-
rounding mean flow. Several studies were carried out to
Pr
t

2 u 2x v 2y w z2 u y v x 2 v z w y 2

quantify the mean flow effects on sound radiation.18,19 It was


found that the mean flow results in not only the refraction of
2
w x u z 2 u x v y w z 2 ,
3 23
the radiated sound, but also convection amplification due to
fluid motion. The Mani et al.11 formulation is used to link the where t is the turbulent viscosity, and (u, v ,w) are the
turbulent properties of a jet plume to the radiated acoustic plume velocity components in the (x,y,z) directions, respec-
field as follows: tively. In the extended version of the (k) turbulence
model20 used in the current CFD computation, the source
p 2 R, ,
V
a xx 4a xy 2a y y 2a yz d 3 r, 17
term in the dissipation rate equation is given by
S C 1 Pr /k C 2 /k C 3 PrPr/k ], 24
with given by which shows the presence of two time scales; k/ and k/Pr.


Therefore, an acoustic model that accounts for radiation from
I 2
C 2
1M cos 2 both time scales is needed. This model determines the far-
16 R
2 2
C 4 C field mean-square-pressure as the average of the mean-
1M c cos 1 , 18 square-pressure given by the individual time scales following

where is the ambient density of the fluid, and C is p 2 12 p 2 p Pr


2
, 25
the local speed of sound. The directivity factors where p 2 and p Pr
2
are associated with 0, and 0,Pr of Eqs.
(a xx ,a xy ,a y y ,a yz ) have different forms depending on the lo- 21 and 22, respectively. The model defined by Eq. 25 is
cation of the source, the velocity, and temperature profiles in referred to as the dual time scale model.
the vicinity of the source. These factors depend explicitly on
a shielding function g 2 that has the form
III. THE CFD DATABASE
1M cos 2 C /C 2 cos2
g 2 r . 19 In order to use the far-field acoustic radiation model
1M c cos 2 derived above, one needs the local steady-state engine plume
Depending on the location and the shape of the profiles i.e., data ( ,C,M ,k,,Pr,U) and the plume grid. If the test con-
velocity and temperature, the function g 2 may have both ditions are known accurately, only one dataset is required.
positive and negative regions in space. When a negative re- However, if more test conditions are planned then one needs
gion exists, fluid shielding of the source is possible. The to have a CFD solution for each set of conditions. In the
directivity factors are given by example presented here, the Reynolds-averaged-Navier
Stokes RANS equations are integrated over a domain 2.5
cos4 times the length of the Venture Star space vehicle of 25 m
a xx , 20a
1M c cos 4 xx this vehicle is also known as the X-33. The CFD code used,
known as FDNS,21 uses a second-order finite difference
g 20 cos2 scheme to discretize the diffusion fluxes and source terms of
a xy xy , 20b
2 1M c cos 2 the governing equations. The convective terms are dis-
cretized using a second-order total-variation-diminishing dif-
a y y 38 g 40 y y , 20c ference scheme. The volume grid is composed of 2.2 million
grid points as shown in Fig. 2. Details about the grid refine-
a yz 18 g 40 yz , 20d
ment techniques and the treatment of the various boundary
where g 20
is the value of g 2 (r) at the source radius rr 0 . conditions and source terms can be found in Ref. 22. An
The shielding coefficients ( xx , xy , y y , yz ) depend on the extended (k) turbulence model describes the turbulence.
location of the source and are set equal to 1 in the present Figure 3 shows the temperature distribution in the plume
study i.e., no shielding of the source. region.
In Eq. 12, a characteristic time delay, 0 , was intro-
duced. The most widely used value for this time scale is IV. EXPERIMENTAL MEASUREMENTS
k NASA conducted ten single-engine tests of the linear
0, c t , 21
aerospike engine, XRS-2200 at the Stennis Space Center in
late 1999 and early 2000. Figure 4 is a photograph of this
where is the dissipation of the turbulent kinetic energy. In liquid-fueled rocket engine and the initial part of the exhaust
this paper, one more time scale is considered plume. The engine was suspended above a J-deflector where
k the vertical distance along the centerline of the engine from
0,Prc Pr
t . 22 the engine exit ramp plane to the deflector was 15.244 m. For
Pr
CFD models, relevant dimensions of the test configuration
The production of the turbulent kinetic energy, Pr, in Eq. 22 include the nozzle ramp 2.822.24 m at one end and 1.07
is written as 2.24 m at the other end; the ramp is 1.27 m long and

2038 J. Acoust. Soc. Am., Vol. 112, No. 5, Pt. 1, Nov. 2002 Frendi et al.: Jet engine noise measurement and prediction
FIG. 2. Computational grid for the X-33 vehicle.

possesses an area ratio of 58. The jet Mach number was


3.042. The test setup included microphones that were ar-
ranged at ground level at various distances and angles from
the test stand, as shown in Fig. 5. The Ms and Ns of
Fig. 5 designate the locations of the microphones and ampli-
fiers. The microphones were vented, polarized capacitor
type, 1.27 cm in diameter B&K 4190 and 4191. The mi- FIG. 4. Picture of the hotfire test showing the water vapor enhanced engine
crophone signals were preamplified B&K 2669 and passed plume.
through shielded cabling to amplifiers Nexus 2690, then
through coaxial cables to an analog tape recorder. A Teac tic output of 94 dB and 1000 Hz: the microphone output
XR-510 tape recorder captured the signal with operational signal for this tone was recorded. The reference pressure
conditions of 5 volts full scale at 76-cm/s tape speed in the used for the decibel level is 20 mPa.
FM mode. Prior to each XLS-2200 test, an end-to-end cali- The microphone frequency range was 5 Hz to 20 kHz,
bration procedure was used. The crew placed a calibrator and full-scale sound levels varied from 140 to 160 dB de-
B&K 4231 over each microphone that produced an acous- pending on location. For each microphone, the background
noise was measured and removed from the test data at these
high noise levels, the background noise was negligible.
These data records were processed after each test and pro-
vide a database for comparison with linear aerospike acoustic
environment predictions. The database includes spectral
shapes, overall sound-pressure-level distributions, and decay
rates.

V. ACOUSTIC MODEL VALIDATION


Initially, the empirical constants of the model

(c c ,c d ,c l ,c Pr
t ,c t ) are selected by using a measured spec-
trum in 1/3-octave frequency bands at a fixed observer loca-
tion. As depicted in Fig. 5, several sound measurements are
available at a radius of 90 m from the XRS-2200 test stand
and at various angular locations with respect to the deflected
plume. Figure 6 shows a comparison between the measured
and calculated spectrum at the 90-deg location. This spec-
trum is obtained after a few trial and error runs with different
empirical constants. In addition to using the experimental
spectrum, the overall sound-pressure level OASPL at the
same location is used in choosing the constants. The final set
of selected empirical constants is c c 0.27, c d 0.55, c Pr t
FIG. 3. Temperature distribution in the plume region near the vehicle ob- 0.04, c t 0.045, and c l 1.13. Figure 7 shows the calcu-
tained by CFD. lated and measured overall sound-pressure level as a function

J. Acoust. Soc. Am., Vol. 112, No. 5, Pt. 1, Nov. 2002 Frendi et al.: Jet engine noise measurement and prediction 2039
FIG. 7. Measured and predicted overall sound-pressure level at various
angular locations in the midfield.

VI. THE MEASUREMENT AND PREDICTION TOOL


The innovative aspect of the proposed measurement tool
resides in its capability to make predictions while taking ex-
perimental measurements and display both the predicted and
measured results in real time i.e., during a test. The acous-
tical engineer, using the tool, can adjust two of the empirical
FIG. 5. Midfield measurement locations around the test stand. constants (c c ,c t ) on the display screen and enter the location
of the microphone so that the semiempirical model can pro-
vide predictions at that particular location. The two empirical
of the angular location. Good agreement is obtained at most constants given above are chosen because the model is more
locations except near 180, where a wall separates the mi- sensitive to these constants than to the other ones. Figure 8
crophone from the engine. This is not taken into account by shows the major functional components of the system. In
the far-field acoustic model. order to perform its task correctly, the tool needs a CFD
database results from prior CFD computations for the jet
engine being tested and at the same conditions, which will
be used in the integration of the far-field acoustic model, and
the experimental data being collected in real time.
These components can be used as stand-alone or inte-
grated. In a stand-alone application, the semiempirical model
will use previously optimized empirical coefficients to make
acoustic radiation predictions for any combination of new
engine test conditions. A CFD plume solution that matches
the new test condition is necessary to complete the model.
Figure 9 shows four screen displays obtained by adjusting
the above-mentioned empirical constants. Each display
shows the experimental data together with the predicted
spectrum.

VII. CONCLUSIONS
A new, innovative measurement and prediction tool has
been designed and tested. The purpose of the tool is to mea-
sure the far-field noise radiated by an engine plume. In ad-
dition, the tool is equipped with a prediction package based
FIG. 6. Comparison of the measured and predicted spectrum at a midfield
location 100 meters away from the test stand and at 90 with the deflected on a semiempirical far-field acoustic radiation model and an
plume. existing CFD database of engine plume. Test results show

2040 J. Acoust. Soc. Am., Vol. 112, No. 5, Pt. 1, Nov. 2002 Frendi et al.: Jet engine noise measurement and prediction
FIG. 8. The various functional components of the pro-
posed measurement and prediction tool.

FIG. 9. Various screen displays of the measured and predicted results using proposed tool.

J. Acoust. Soc. Am., Vol. 112, No. 5, Pt. 1, Nov. 2002 Frendi et al.: Jet engine noise measurement and prediction 2041
10
that the tool is able to display both the predicted and mea- C. K. W. Tam, Supersonic Jet Noise, Annu. Rev. Fluid Mech. 27,
sured results in real time. The semiempirical model has been 17 43 1995.
11
R. Mani, E. J. Stringas, J. C. F. Wang, T. F. Balsa, P. R. Gliebe, and R. A.
tested and shown to give good agreement with far-field mea-
Kantola, High Velocity Jet Noise Source Location and Reduction, Task
surement. 2, FAA-RD-76-79-II 1977.
12
J. E Ffowcs Williams, The Noise from Turbulence Convected at High
ACKNOWLEDGMENTS Speed, Philos. Trans. R. Soc. London, Ser. A 255, 469503 1963.
13
A. Khavaran, E. A. Krejsa, and C. M. Kim, Computation of Supersonic
A. Frendi, Wade D. Dorland, and Thein Maung would Jet Mixing Noise from an Axisymmetric CD Nozzle Using (k) Tur-
like to acknowledge the support of NASA Marshall Space bulence Model, AIAA Paper 92-0500, Jan. 1992.
14
Flight Center under an SBIR Phase I Grant No. ASRI-NAS8- C. Baily, W. Bechara, P. Lafon, and S. Candel, Jet Noise Predictions
01009-01 with Tom Nesman as the technical monitor. Using a (k) Turbulence Model, AIAA Paper 93-4412, Oct. 1993.
15
C. Baily, P. Lafon, and S. Candel, Prediction of Supersonic Jet Noise
1
from a Statistical Acoustic Model and a Compressible Turbulence Clo-
A. Frendi, On the CFD Support for the Hyper-X Aerodynamic Data- sure, J. Sound Vib. 1942, 219242 1996.
base, AIAA paper 99-0885 1999. 16
C. K. W. Tam and L. Auriault, Jet Mixing Noise from Fine-Scale Tur-
2
A. Frendi, T. E. Nesman, and T.-S. Wang, Computational and Experi- bulence, AIAA J. 372, 145153 1999.
mental Study of the Linear Aerospike Engine Noise, AIAA J. 398, 17
G. K. Batchelor, The Theory of Homogeneous Turbulence Cambridge
14851492 2001.
3 University Press, Cambridge, 1953.
M. J. Lighthill, On Sound Generated Aerodynamically. I. General 18
T. F. Balsa, The Acoustic Field of Sources in Shear Flow with Applica-
Theory, Proc. R. Soc. London, Ser. A 211, 564 587 1952.
4 tion to Jet Noise: Convective Amplification, J. Fluid Mech. 791, 33
H. S. Ribner, Quadrupole Correlations Governing the Pattern of Jet
1977.
Noise, J. Fluid Mech. 381, 124 1969. 19
5
H. S. Ribner, On the Strength Distribution of Noise Sources Along a M. E. Goldstein, High Frequency Sound Emission from Moving Point
Jet, J. Acoust. Soc. Am. 31, 245246 1958. Multipole Sources Embedded in Arbitrary Transversely Sheared Mean
6
H. S. Ribner, On Spectra and Directivity of Jet Noise, J. Acoust. Soc. Flows, J. Sound Vib. 804, 499 1982.
20
Am. 35, 614 616 1963. Y.-S. Chen and S. W. Kim, Computation of Turbulent Flows Using an
7
J. E. Ffowcs Williams, Hydrodynamics Noise, Annu. Rev. Fluid Mech. Extended k Turbulence Closure Model, NASA CR-179204, Oct.
1, 197222 1969. 1987.
8 21
J. E. Ffowcs Williams, Aeroacoustics, Annu. Rev. Fluid Mech. 9, 447 Y.-S. Chen, FDNSA General Purpose CFD Code, Version 4.0, Engi-
468 1977. neering Sciences, Inc., ESI-TR-97-01, Huntsville, AL, May 1997.
9 22
M. E. Goldstein, Aeroacoustics of Turbulent Shear Flows, Annu. Rev. T.-S. Wang, Analysis of Linear Aerospike Plume-Induced X-33 Base-
Fluid Mech. 16, 263285 1984. Heating Environment, J. Spacecr. Rockets 366, 777783 1999.

2042 J. Acoust. Soc. Am., Vol. 112, No. 5, Pt. 1, Nov. 2002 Frendi et al.: Jet engine noise measurement and prediction

View publication stats

You might also like