2008-28-0059

Numerical Analysis of Sunroof Buffeting

J. S. Rao, M. Saravana kumar and Ashish Singh
Altair Engineering (India), Bangalore 560 087

Copyright © 2008 SAE International

ABSTRACT directivity patterns of these sources are shown in Fig. 1,

which gives resemblance to the transmission direction.
Noise computations are carried out for sunroof opening
in automotives at a resonating speed. First, the behavior
of baseline opening shape is analyzed through Lighthill introduced the notion “analogy” of representing
frequency spectrum obtained from acoustic analysis. a complex fluid mechanical process that acts as an
Then the opening shape is changed to reduce the sound acoustic source by an acoustically equivalent source
pressure level of buffeting noise. The analysis with new term. The notion of “analogy” refers here to the idea of
shape is able to reduce the Noise dB levels up to 5dB. representing a complex fluid mechanical process that
Commercial software, Fluent, is used to compute the acts as an acoustic source by an acoustically equivalent
turbulent flow field. Large-Eddy Simulation is used for source term. While Lighthill’s equation is formally exact
better prediction of fluctuating components. Acoustic (i.e. derived without approximation from the Navier-
results obtained from Ffowcs-Williams and Hawkings Stokes equations), it is only useful when we consider the
equations are well compared with other’s work. case of a limited source region embedded in a uniform
stagnant fluid. At least we assume that the listener which
INTRODUCTION detects the acoustic field at a point x at time t is
surrounded by a uniform stagnant fluid characterized by
Sound can be measured as a pressure perturbation in a a speed of sound c0. Hence the acoustic field at the
medium. The study of sound is known as Acoustics, and listener should accurately be described by wave
when the sound is generated by air flow, the study equation (1). The key idea of the so-called “aero-
becomes Aero-acoustics. acoustic analogy” of Lighthill is to derive from the exact
equations of motion a non-homogeneous wave equation
Analytical solutions for the acoustic wave equations with the propagation part as given by equation (1).
governed by flow are complicated hence computational Hence the uniform stagnant fluid with sound speed c0,
Aero-Acoustics is intensively developed for the noise density 0 and pressure p0 at the listener’s location is
computations in recent decade [1]. This area includes assumed to extend into the entire space, and any
the sophisticated flow field calculation and acoustic departure from the “ideal” acoustic behavior predicted by
signal processing (sound source, transmission, equation (1) is equivalent to a source of sound for the
dissipation and Fourier transform) for the prediction of observer. The following equation is a general form of
noise levels generated by turbulent fluid flow alone or acoustic equation derived from N-S equation considering
fluid solid interaction. the effect of moving surfaces too. Except the last two
terms on right hand side, the equation is known as
The production of noise is highly dependent of sound Lighthill’s acoustic equation. Ffowcs-Williams and
sources, which may be categorized in the following Hawkings [2, 3] introduced the effect of surfaces which
types, covers all type of sources described above.
1. Monopole
2. Dipole w2U ' w2U ' w 2Tij
3. Quadrupole  c 02
wt 2 w x i wx j wx i w x j
A monopole represents volume velocity or volume
w ª ' wf º w ª wf º
displacement sources. A Dipole is a pair of closely  «( p G ij ) G ( f )»  « U 0ui G ( f )»
spaced Monopoles; a point force is directed along the wx i «¬ wx j »¼ wt ¬ wx i ¼
Dipole axis. While a Monopole can be considered as a
Mass source, a Dipole is a Force source. Fluid-Solid
interaction and boundary layer turbulence are typically (1)
Dipole like. A point Quadrupole is an array of two
dipoles, there is no net force and it exerts an oscillating where
moment as in oscillatory rotation in turbulent flows. The
409
 c A
Tij Pij  Uu i u j   c02 U ' p0 G ij fH
2S VL
 
Uu i u j  W ij  p ' c02 U ' G ij
(2) (3)

In equation (1) the right hand side terms in order are where, A is the area of the opening, L, length of the
neck, V, volume of the cavity or cabin and c is the
1. Tij ~ Lighthill’s stress tensor which contains velocity of the sound. It can be observed that larger
velocity variations related acoustic source terms opening area gives higher frequency since air can rush
equation (2) (Quadrupole) and entropy acoustic in and out faster. Larger volume gives lower frequency
source terms since more air must move out to relieve a given pressure
excess. Also larger neck gives lower frequency because
2. Dipole sources are due to pressure variations at of more resistance to air moving in or out.
physical solid surface
The physics of generation and convection of vortices
3. Monopole sources are due to displacements of from the shear layer along the sunroof can be explained
fluid by accelerating/decelerating surfaces. in the following way. As the self-sustained oscillations
inside open cavity (sunroof) become prominent at certain
Correct prediction of these source terms is dependent on range of flow velocities, vortices are shed at the leading
flow field results accuracy. RANS equations are time edge of cavity opening and continue to be convected
averaged hence miss the accountability of fluctuations. along the opening length as pressure inside cabin
Large Eddy Simulation LES is a transient turbulence changes. These vortices are found to be convected with
model that falls midway between RANS and DNS convective velocity half of the mean flow velocity [5, 6].
models. Large eddies contain most of the turbulent At the rear edge of opening, these vortices break and
kinetic energy; scale sizes are on the order of the flow generate pressure wave which enters in the cavity
passages. The energy cascades from large to small generating high amplitude noise for driver and
eddies and small eddies dissipate the energy they passengers who are sitting inside the resonant body
receive from larger eddies in the spectrum. In acoustics itself. Based on convective velocity of hydrodynamic
studies large eddies are captured in the flow which act instabilities, another formulation for vortex shedding
as sources for sound propagation. frequency can be given as

Large eddy simulation is based on space-filtered fs U f / 2l

equations; time dependent calculations are performed
and large eddies is explicitly calculated. For small (4)
eddies, their effect on the flow pattern is taken into
account with a sub grid scale (SGS) model of which where U is the free stream velocity and l is the spacing
many styles are available. A simple sub grid scale model between two convective vortices. Equation (4) can be
is based on Bossinesq approximation in which sub grid approximated from general formulation for shedding
scale stress tensors are modeled as turbulent viscosity frequency based on number of eddies n present at
multiplying with the resolved stress tensor. Mathematical opening area.
derivations and formulations for large eddy simulations
are discussed in length in [4].
fs 1 / 3(n  1 / 4)U f / l
SUNROOF BUFFETING
(5)
Aerodynamic flow over an open sunroof generates
significant noise inside the automotive vehicle’s cabin. There are many expressions available to estimate the
Sunroof buffeting is the practical example of Helmholtz vortex shedding frequency for the shear layer. Above
resonator. When a cavity, see Fig. 2, is subjected to a equation is one of them for turbulent boundary layer,
continual grazing flow, it produces strong periodic where U is the free stream velocity, l is the length of the
oscillations in the interior of the cavity region at various sunroof opening, n is the number of vortices present
flow velocities. It results in self sustained oscillations at across the opening.
the Helmholtz acoustic resonance frequency. In the case
of an automobile, the open sunroof acts as the open If it is assumed that the vortices are being shed regularly
cavity which under the influence of a grazing flow causes and being carried by main stream velocity in
the Helmholtz resonance frequency of the cabin to be downstream direction, the frequency of vortex shedding
excited [5, 6]. Helmholtz mode of a cabin is a special can be calculated from above equation. The frequency
case in which a standing wave forms with a distinctly prediction from vortex shedding can be very near to
lower frequency and high amplitude. The resonant actual but not a correct one because vortex shedding is
frequency of a Helmholtz resonator can be given by not the only reason for noise generation in sunroof
following formulation. buffeting. It is the feedback mechanism [7, 8] of
410
downstream convecting hydrodynamic instabilities and FLOW FIELD RESULTS
upstream convecting acoustic waves due to rear edge
reflections, which contributes in high amplitude noise This section gives the flow field results for suppressing
tone. This is known as buffeting noise and efforts have sunroof buffeting noise [9]. The qualitative difference in
been made towards the prediction of frequency and the transient snap shot of dynamically steady state
amplitude of this tone. This buffeting noise becomes conditions is shown for both the baseline and optimized
dominant when resonance occurs due to similar cases. Computationally the dynamic steady state is
Helmholtz and vortex shedding frequency. reached at twice of the flow traveling time in domain
(32D/Usec). Fig. 4 shows the contours of dy/dx and
NUMERICAL ANALYSIS static pressure for baseline and optimized cases. Sharp
corner at front edge of opening excites the instability and
In the present work the aero-acoustics analysis is feeds strong vortices continuously, which convect
performed to capture the physics of sunroof buffeting downstream and hit sharp rear edge. This causes strong
and to estimate the sound pressure level from the vortex breakdown, which pulsates the pressure waves
frequency spectrum. Fluent 6.2 is used to combine inside the car body in baseline geometry and cause
computational aero-acoustics (CAA) with computational unwanted noise for the passengers. On the other hand
fluid dynamics (CFD) for the prediction of acoustics rounded front edge reduces the excitation of instability
signals from source to receiver. Simple geometry of a and intensity first and while interaction of instability with
sedan is taken with sunroof. The shape of opening is rounded rear edge also suppresses the vortex
changed through morphing to meet the noise breakdown phenomenon so the generated pressure
suppression requirement. The dimensions of the pulses are less hence noise too. The bent in the rear
baseline geometry are listed in Table 1. Fig. 3a shows edge also helps in passing the vortices downstream
the computational domain, 3b gives the mesh and without much solid obstruction. The above explanation is
boundary conditions and 3c receiver location. The qualitatively analyzed by comparing the flow field
geometry of the sun roof is shown in Fig. 3d. contours in Fig. 4. A number of vortices in the opening
area are captured, see Fig. 5. Strong vortex generation
The fluid domain behind the car is extended to 23D (Fig. and convection is found in baseline geometry with four
3a). The reason for such extended domain is to avoid vortices Vo1 to Vo4 at a particular time instant.
the back flow, which usually occurs due to rotating,
eddies at outlet boundary. This back flow may lead to a The low pressure crest in the plot shows the vortex
wrong prediction of sources and hence acoustic signals. center in the physical domain. Quantitatively pressure
The outer domain is subjected to non-reflecting crests in baseline case are higher than optimized case
boundary conditions, which removes the effect of near to rear edge. This can be explained with vortex-
reflection from walls. Velocity inlet and pressure outlet structure interaction which is more in baseline case
boundary conditions are used in the problem. Meshing of because the vortex breakdown is more compared to
the fluid domain is done carefully with all quad elements. optimized case and hence the higher center pressure.
Two major factors for quad mesh generation are the type On the other hand in optimized case the vortex is
of turbulence model used and the smooth boundary keeping its existence with little distortion near rear edge
layers on solid surfaces, which are major sources of and hence not surprisingly adding more pressure to the
noise generation. Dense mesh is taken in the region of center. Due to rounded front edge and distortion of
opening to assure good quality of acoustic signals. Total vortex in optimized case, the total number of vortices is
number of elements is 40,000. Time step is set only three Vo1 to Vo3 at the same time instant.
according to equation (6) and the update in solution is
commanded with non iterative time advancement. ACOUSTIC RESULTS

Gt CGx / u local The flow field data obtained from computation is used for
the aero-acoustic analysis. The only means to get the
(6) sound pressure level (amplitude) is to do the spectral
analysis. It requires the transformed source signals
To analyze acoustic signals, a receiver point (Fig 3c) at through Ffowcs William and Hawkings equation in the
driver location is specified since the cabin is supposed to form of acoustic signals. The Fast Fourier Transform
affect by buffeting noise. The free stream velocity is (FFT) of the acoustic signals at the receiver location
taken 25m/s, which is close to bringing the resonance produces frequency spectrum where the Sound
condition with the car cavity. After the baseline analysis, Pressure Level is analyzed for different tones in sunroof
different geometries for the sunroof are studied using buffeting. In the current analysis the driver location is set
morphing in Hypermesh and the one that gives a as a receiver point and the signals are analyzed for
substantial reduction in the noise level is denoted baseline and optimized case. The run is extended for
optimized geometry, see Table 1 and Fig. 3d. double the traveling time and flow field simulations are
entered as input to computational aero-acoustic
analysis. The FFT spectrum Fig. 6 obtained from
baseline run gives a SPL peak at 98dB corresponding to
68Hz frequency. Other high modes are also present at
411
different frequencies. The highest mode corresponds to 8. Henderson, B., “Category 6, Automobile Noise
resonance condition between vortex shedding and Involving Feedback-Sound generation by Low
Helmholtz frequency. Spectrum shows that apart from Speed Cavity Flows”, Third Computational
the turbulent broadband noise there is a recognizable Aeroacoustics (CAA) Workshop on Benchmark
signal from sunroof buffeting. The SPL level is nearly problems, NASA CP-2000-209790, August, 2000.
10dB low compared to others [10] and this difference is 9. Ota, D. K., Chakravarthy, S. R., Becker, T and
attributed to two dimensional simulations taken for Sturzenegger, T., ``Computational Study of
present work rather than three dimensional works in [8]. Resonance Suppression of Open Sunroofs’’, Journal
Similar run for optimized case produces more like of Fluids Engineering, vol. 116, pp 877-882, 1994.
broadband signals in Fig. 7. No separate high amplitude 10. Singh, R, ``Sunroof Buffeting Simulations Using
tone is figured out here. The noise level is reduced by
Pam-FLOW’’, Ameripam-2001, User’s Conference,
5dB approximately for the same resonance frequency.
Nov. 2001.
This reduction in noise level is related to low vortex
breakdown at the rear edge of opening.
CONTACT
CONCLUSION
Professor J. S. Rao, ASME Fellow, is currently Chief
Science Officer, Altair Engineering, Bangalore. He is
An acoustic analysis is performed for sunroof opening
also President of The Vibration Institute of India, and
noise at resonance condition. The flow physics is
Director on the Board of GMR Energy Group, Bangalore.
explained through analysis and the frequency spectrums
Dr. Rao obtained his Ph D (1963) and D Sc (1971) from
for baseline and optimized case are compared. Current
IIT Kharagpur; He has been chairman and BHEL Chair
work is able to reduce the noise level up to 5dB for
Professor in Mechanical Engineering at IIT Delhi and
optimized case. The baseline noise tones are compared
Visiting Professor at University of Surrey, National
well with others work. The optimized shape is generated
Institute of Applied Sciences, Lyon, Concordia
through morphing on Hypermesh. Further work on
University, Rochester Institute of Technology, Technical
optimization demands a synchronized development of
University, Kassel, National Chung Cheng University,
flow filed and prediction of higher tones. This will be
Chia-Yi, The University of New South Wales, Sydney.
achieved though an optimization study on Hyperstudy
He has been a Consultant to GTRE, BHEL, ISRO in
along with the simultaneous run on Fluent.
India, Washington Metro and Stress Technology Inc., in
USA and Twyflex Couplings in England, Institute of Fluid
ACKNOWLEDGMENTS Flow Machinery Polish Academy of Sciences, Taipower
Company Taipei etc. He successfully led a team of
The authors are grateful to Altair Engineering India Pvt BHEL engineers in investigating the complete
Ltd for their support in this work. catastrophic failure of Narora Atomic Power plant; fixing
up annoying squeal problem of Washington Metro trains,
REFERENCES identifying failure problems of High pressure compressor
rotor failures of Kaveri Engine and advanced design of
1. Karbon, K. and Kumarasamy, S., ``Computational high speed cryo engines in space application among
Aeroacoustics Applications in Automotive Design’’, others. He is a founder member of the IFToMM, member
First MIT Conference on Computational Solid and of its executive council and Chairman of its Rotor
Fluid Mechanics, MIT, June 2001. Dynamics Committee and Permanent Commission on
2. Ffowcs Williams, J. E., Hydrodynamic Noise, Annual Conferences, member of History of Machines and
Review of Fluid Mechanics 1, p 197-222, 1969 Mechanisms technical committee, Chief Editor of the
3. Fluent 6.3 documentation and help files, Acoustic journal “Advances in Vibration Engineering”. He
1.0 module documentation and help files authored over 350 research publications and 10 text
4. Boris, J. P., Grinstein, F. F., Oran, E. S. and Kolbe, books, guided 31 students for doctoral degree. He has
R. J., “New Insights into Large Eddy Simulation”, conducted over 25 short courses on Turbomachine
Fluid Dynamics Research, vol.10, No.4-6, pp.199- Blade Vibrations and Lifing and Rotor Dynamics and
227, 1992. Condition Monitoring, in USA, Germany, Canada,
5. Nelson, P. A., Halliwell, N. A., and Doak, P. E., Australia, Singapore, Taiwan, Venezuela, Korea, Hong
Kong, Malaysia, Korea and Dubai. He has been honored
``Fluid Dynamics of Flow Excited Resonance, Part I:
by several scientific bodies, including Pacific Center of
Experiment’’, Journal of Sound and Vibration, vol. 78
Thermal-Fluids Engineering, Hawaii and The
(1), pp 15-38, 1981.
International Federation for Promotion of Mechanism
6. Nelson, P. A., Halliwell, N. A., and Doak, P. E., and Machines. Professor Rao is a past president of The
``Fluid Dynamics of Flow Excited Resonance, Part II: Indian Society of Theoretical and Applied Mechanics and
Flow Acoustic Interaction’’, Journal of Sound and elected life time Emeritus Chairman of IFToMM
Vibration, vol. 91(3), pp 375-402, 1983. Technical Committee on Rotor Dynamics.
7. Rienstra, S. W., Hirschberg, A., “An Introduction to
Acoustics”, IWDE 99-02, Eindhoven University of e-mail ID: js.rao@altair.com
Technology, 1999.

412
 Fig. 1 Sound Sources

Flow

A
L

Fig. 2 Helmholtz cavity

6D

5D 4D 23D

L
A

D
V

Fig. 3a Computational domain

Symmetry

Velocity inlet Pressure outlet

Symmetry

Fig. 3b Mesh and boundary conditions

413
 Receiver

Fig. 3c Receiver location

Fig. 3d Baseline and Optimized shape

Table 1 Opening area dimensions

Case Opening area Neck length, radius Free stream velocity

Baseline 0.26 m2 0.026 m, 0 m 25 m/s
Optimized 0.26 m2 0.026 m, 0.014 m 25 m/s

414
 Fig. 4 Contours based on flow field data

Fig. 5 Convecting eddies in the opening area for baseline and optimized case

415
Fig. 6 Frequency spectrum for baseline case

Fig. 7 Frequency spectrum for optimized case

416