Professional Documents
Culture Documents
In recent years, as engineering design of components and includes cases where the frequency range of interest is fair-
systems has become increasingly sophisticated, a signif- ly narrow, the sources and receivers are located close to
icant amount of effort has been directed toward the each other, and the sound to be captured is fairly loud. The
reduction of aerodynamically generated noise. With the sound generated by an open car window (see page 5) is one
ongoing advances in computational resources and algo- example, and the sound produced by a side view mirror is
rithms, CFD is being used more and more to study another [1, 2]. For both of these cases, the CAA results are
acoustic phenomena. Through detailed simulations of in good agreement with experimental measurements.
fluid flow, CFD has become a viable means of gaining
insight into noise sources and basic sound production CAA has also been used successfully to predict whistles
mechanisms. (loud tones) produced by
automotive air intake
FLUENT offers four systems. The whistling
approaches for simulating sound is caused by an air
aeroacoustics. In order of jet passing underneath
decreasing computational the throttle plate (Figure
effort, these are computa- 1). As it passes over a
tional aeroacoustics sump cavity, a shear
(CAA, or the direct layer is established. If
method), the coupling of resonance occurs
CFD and a wave-equa- between the flapping
tion-solver, integral shear layer and sound
acoustic models, and Figure 1: Instantaneous velocity magnitude contours in a CAA simulation waves bouncing off the
broadband noise source of whistles generated in an automotive air intake system; the shear
layer flapping in the mouth of the sump cavity causes a loud whistle
sump bottom, a loud
models. whistle develops. The
sound spectrum predict-
Computational Aeroacoustics ed by a CAA simulation (Figure 2) is in excellent agree-
Computational aeroacoustics is the most comprehensive ment with the corresponding experimental measurement
way to simulate aeroacoustics. It does not rely on any [3, 4]. The CAA simulation predicts almost the exact
model, so is analogous to direct numerical simulation same whistle frequency and sound pressure level (SPL)
(DNS) for turbulent flow. CAA is a transient simulation as measured in the experiments.
of the entire fluid region, encompassing the sources,
receivers, and entire sound transmission path in between.
By rigorously calculating time-varying flow structures,
pressure disturbances in the source regions can be fol-
lowed. Sound transmission is simulated by resolving the
pressure waves traveling through the fluid. While CAA is
the most general and accurate theoretical approach for
simulating aeroacoustics, it is unrealistic for most engi-
neering problems because of a number of practical limi-
tations, including widely varying length and time scales
characteristic of the sound generation and transmission
phenomena, and widely varying flow and acoustic pressures.
References:
1. R. Siegert, V. Schwarz and J. Reichenberger, AIAA
Paper No. 99-1895.
2. B.S. Lokhande, S.D. Sovani and J. Xu, SAE Paper
No. 2003-01-1698.
3. V. Kannan, J. Seifert, T. Golletti and D. Hanner,
Figure 6: The Direct (CAA) and FW-H approaches are both in good SAE Paper No. 2004-01-0395.
agreement with experiment for a receiving point not far from the mirror 4. V. Kannan, S.D. Sovani, D. Greeley and A.D. Khondge,
Data courtesy of DaimlerChrysler
Submitted to SAE NVH Conference, May 2005.
Broadband Noise Source Models 5. M.J. Lighthill, Proc. Royal Society A 211, p. 564 (1952).
The three methods described so far require well-resolved 6. J.E. Ffowcs-Williams and D.L. Hawkings, Proc.
transient CFD simulations, since they aim to determine Royal Society of London A 264, pp. 321-342 (1969).
the actual time-varying sound-pressure signal at the 7. Revel, Lockheed Report 28074.
receiver, and from that, the sound spectrum. In several 8. G.M. Lilley, The Radiated Noise from Isotropic
practical engineering situations, only the locations and Turbulence Revisited, NASA Langley Research
relative strengths of sound sources, rather than the sound Center ICASE Report 93-75; NASA CR-191547.
spectra at the receivers, need to be determined. If the
sound is broadband (without any prominent tones charac-
terized by sharp peaks in the spectrum), the source
strengths can be evaluated with reasonable accuracy
from the time-averaged structure of the turbulent flow in
the source regions.
Wind buffeting, the noise and pulsating forces that are At DaimlerChrysler, engineers have been using the com-
experienced when driving a car with the side windows putational aeroacoustics (CAA) approach in FLUENT to
open, has become a significant factor in the overall pas- simulate wind buffeting. In two recent studies [1, 2], the
senger experience in recent years. It is caused by an modeling process began by importing a surface model of
unstable shear layer that is established at the upstream the outer shape of the vehicle and a CAD model of the
edge of the window opening. Disturbances are shed from vehicle interior (Figure 1) into a CFD preprocessor. The
this location and travel along the side of the vehicle. simulation domain was defined to include the entire pas-
When they reach the rear edge of the window opening, a senger cabin, which is connected to the external flow
pressure wave is generated that propagates both inside domain through an open window. Dummies representing
and outside the passenger compartment. Outside the the passengers were included in the model to correctly
vehicle, this wave propagates both forward and backward represent the volume of the passenger compartment. The
along the side of the car. When the forward traveling vehicle surface was modeled to a significant degree of
wave reaches the front edge of the opening, it triggers detail to capture flow development from the vehicle front
another disturbance that moves back to the rear edge. end to the window opening. Several levels of local mesh
This process is repeated many times every second and refinement were used. Two refinement levels were
causes the shear layer to develop a characteristic buffet- applied outside the vehicle to capture the wake behind
ing frequency, which depends on the speed of the auto- the vehicle. One refinement level was applied inside the
passenger compartment to capture wave propagation
inside the cabin. The finest refinement level was applied
at the area of the opening to capture the shear layer. A
close-up view of the surface mesh near the open window
is shown in Figure 2.
The turbulent flow was captured using the RNG k-ε and
LES turbulence models, both of which have been shown
in the past to provide good results for a range of turbu-
lent conditions. Interior surfaces of the vehicle were
assumed to be solid walls instead of soft surfaces such as
Figure 1: The model geometry of the car exterior and interior with carpeting or fabric. Actual car surfaces are less reflective
the front left window open
and more absorbent than solid walls, giving the model a
mobile and the geometry of the opening. Often the frequency
is below the range that can be heard by human ears but it
still can be felt by passengers as a pulsating wind force.
Wind buffeting can be detected using microphones, but
the complicated pressure waves that are its cause are very
difficult to measure. As a result, engineers in the past
have had to wait until relatively late in the design process
when prototypes become available to measure this phe-
nomenon. These measurements typically give them little
or no information about what areas of the design are
affecting wind buffeting and what could be done to
reduce it. The only option is to modify and test the proto-
types to see whether individual changes have any effect.
This process is so costly and time-consuming that it is dif-
Figure 2: The surface mesh detail for the car exterior and interior
ficult to identify changes that will improve the design. in the vicinity of the open front window
5"
tendency to overpredict the wind and propagation of these waves
buffeting phenomena. can help engineers understand
exactly how wind buffeting occurs
A steady-state solution was first in a particular design, and can help
obtained for each case, and these them iterate quickly to an
results were used as initial condi- improved design.
tions for a subsequent series of
transient simulations that were The frequency spectra and sound
used to capture the pressure fluc- pressure level were in good agree-
tuations in the vicinity of the Figure 3: Pressure field at a speed of 60 mph ment for all locations studied
open window. Monitors were set and yaw angle of 5 degrees, showing the wakes within the vehicle (Figure 4).
behind the A-pillar and mirror
up at the driver’s and front seat Additional simulations correlated
passenger's ear locations and well with experimental measure-
static pressure was recorded at ments in predicting reductions in
these locations at every time the SPL and frequency of the
step. The initial transients died sound from an open front window
down and the pressure traces as the vehicle speed is reduced
reached a dynamically steady from 60 mph to 50 mph.
solution in roughly 300 to 500 Simulations with the left front win-
time steps. Subsequently, pressure dow wide open and the right rear
traces were recorded for time window open 1 inch were also per-
periods between 1.0 and 2.0 sec- formed. These showed that buffet-
onds - long enough to obtain a ing was substantially reduced.
sound pressure spectrum. The Simulations with a modified side
pressure signals were converted Figure 4: Spectrum of the side window buffeting mirror design reduced buffeting by
to the sound frequency spectrum sound heard by a car driver 13 dB, which also correlated well
by taking a discrete Fourier with experimental measurements.
transform using a Hanning window filter. The sound DaimlerChrysler engineers are making use of these results
pressure level (SPL) was finally converted to dB units. by simulating other vehicles, evaluating the influence of
more parameters, and evaluating different modeling tech-
The CFD predictions were validated by comparing them niques. As simulation is more fully integrated into the
to experimental measurements conducted in a wind tun- design process, this approach should make it possible to
nel. The CFD simulations accurately predicted buffeting substantially reduce wind buffeting in the future. !
frequency and sound pressure level, and matched SPL Courtesy of DaimlerChrysler
and frequency variation trends observed in the experi-
ments. Contours of pressure were examined on two hori- Reference:
zontal planes in the critical front window area (Figure 3). 1. D. Hendriana, S.D. Sovani, M.K. Scheimann, “On
The results indicated that a vertical vortex occurs behind Simulating Passenger Car Window Buffeting,” SAE
the A-pillar (the structural member between the wind- Paper No. 2003-01-1316 (2003).
screen and the front window). An animation of the tran- 2. C.-F. An, S.M. Alaie, S.D. Sovani, M.S. Scislowicz,
sient solution showed vortex movement with the local K. Singh, “Side Window Buffeting Characteristics of
flow, with impingement on the B-pillar (the structural a SUV,” SAE Paper 2004-01-0230 (2004).
member between the front and rear windows). The wave
generated at the B-pillar was shown to propagate into the
passenger compartment. Visualization of the formation
!6
Cavity Noise Generation
A recent survey by J.D. Power and Associates [1] indicates cost. A more economical approach makes use of LES to
that excessive wind noise is a major concern for automo- compute the time-varying pressure field (the noise
bile passengers. Wind noise is generated by features on sources), and a simple acoustic analogy to compute the
the outer body of a car, such as roof-racks, door-gaps, sound transmission. This second approach has been
and side-view mirrors. One prominent feature contribut- applied to the rain gutter, using the Ffowcs-Williams and
ing to the overall wind-noise level is the A-pillar rain gut- Hawkings [2] acoustics model to compute the sound trans-
ter. This rain gutter serves the purpose of collecting and mission. The goals of the simulation are to determine:
draining rain water that would otherwise be swept from 1. the transient flow structure around the rain-gutter
the windshield, past the A-pillar, and onto the side win- 2. the distribution of the pressure coefficient, Cp, along
dows, thereby reducing the visibility through them. It has the base plate
the shape of a long narrow channel spanning the length 3. the frequency spectrum of the static pressure at a point
of the A-pillar and facing the wind. From an aerodynam- on the base plate downstream of the rain gutter, and
ics perspective, the rain-gutter acts as a turbulator, creat- 4. the frequency spectrum at a point above the rain gutter.
ing a highly turbulent flow in its wake. Pressure fluctua-
tions created by such turbulence contribute to wind noise. The results from the CFD simulations are compared to
experimental and computational data reported by
Kumarasamy and Karbon [3].
Figure 6: Sound spectrum at the microphone located above the rain gutter 9"
Low Noise Landing
Landing gear noise is not the first thing that comes to Predicting aeroacoustic noise is not a trivial matter. Only
mind when thinking about noise pollution at a busy air- a minute fraction of the kinetic energy present in the pri-
port. The once dominant jet engine noise has been mary flow is converted into acoustic energy and radiated.
reduced significantly over the past thirty years, primarily To correctly capture the acoustics, the turbulent flow
through the introduction of high bypass turbofan engines. must be calculated with high fidelity. Since turbulence is
As a result, airframe noise has emerged as a leading com- an inherently unsteady phenomenon, a time-consuming
ponent of aircraft noise during the final approach phase transient simulation is required. The large eddy simulation
of a landing. Environmental concerns and noise certifica- (LES) turbulence model with the Smagorinsky subgrid
tion regulations are therefore causing aircraft manufac- scale model was used for the landing gear calculation.
turers to take a closer look at this phenomenon.
Integral techniques that predict the far-field acoustic signal
The main contributors to airframe noise in a landing con- using source data input from a near-field CFD simulation
figuration are high-lift devices, such as slats and deployed have emerged as a promising and economic way to compute
flaps, and surprisingly, the landing gear. Measurements sound levels. The Ffowcs-Williams and Hawkings (FW-H)
have shown that these components are not equally important approach [4], the most universal and complete integral
on all aircraft. method available
While the high-lift today, was used for
devices are noisier the simulation.
on medium size air-
craft, the landing The CAD (STEP)
gear is becoming the model provided
dominant source on by NASA Langley
large airplanes, such was cleaned up in
Figure 1: Vortical structures visualized
as the Boeing 777. using iso-surfaces of the second invari- GAMBIT, and a
ant of the deformation tensor, colored by computational
velocity magnitude
Landing gear sys- grid was built
tems have complex, non-streamlined geometries, and gen- using GAMBIT and TGrid. Boundary layer prisms were
erate highly turbulent wakes. Vortices shed from one com- grown in TGrid, so that the prism cap surface mesh could
ponent impinge on other elements, generating noise with a be used to control the growth and continuity of the tetra-
broad spectrum, from a few hundred Hz to several kHz. If hedral elements away from the boundary layer. Size func-
noise can be predicted using engineering software, modi- tions were used to cluster elements in the vicinity and wake
fications such as fairings and streamlining can be intro- of the landing gear, resulting in a 5.3 million cell mesh, suit-
duced during the design phase. able for an LES simulation.
At Fluent, engineers have recently analyzed a 1/10th The landing gear case was run incompressibly (a valid
scale landing gear model, representative of the gear used approximation for compact sound sources) for nearly
on a Boeing 757 aircraft. The same configuration has 10,000 time steps, or one flow-pass through the domain,
been studied using CFD by researchers at Penn State [1] before the turbulence statistics were sufficiently stabi-
and NASA Langley [2, 3], and will also be tested in a lized and the acoustic source data sampling could be start-
wind tunnel at the Quiet Flow Facility at the Langley ed. The acoustic source data was extracted directly on the
Research Center. The four-wheel landing gear assembly landing gear surface over approximately one additional
contains all of the major components, including the oleo- flow-pass, and then processed with the FW-H solver.
strut, axles, connecting blocks, diagonal struts, a door,
and additional parts that hold the configuration together. The FW-H tool is ideal for predicting far-field radiation in
A flat plate simulates the wing surface. the absence of external scattering surfaces. The necessary
source data can be extracted from permeable (interior) or
! 10
Figure 2: Dipole source strength,
using contours of dp/dtRMS, shows
a high source intensity at the rear
diagonal strut and behind the
oleo-strut
Fluent Sweden AB
Göteborg, Sweden
Tel: 46 31 771 8780
Email: info@fluent.se
©2005 Fluent Inc. All rights reserved. FLUENT is a registered trademark of Fluent Inc. All other products or brands are trademarks of their respective holders.