1American Institute of Aeronautics and Astronautics

SIMULATION AND DESIGN OF AUTOMOBILE SUNROOF BUFFETING NOISE CONTROL

Kenneth J. KarbonGeneral Motors Corporation, Warren, MI, USA

Rajneesh SinghAltair Engineering Inc, Troy, MI, USA

ABSTRACT

This paper presents the success story of an application of Computational Fluid Dynamics (CFD) analysis for automotive sunroof buffeting simulation and noise control design. Computational analyses of flow over an open sunroof of a car are performed to study the buffeting phenomenon and to determine the buffeting noise magnitude and frequency. Computations are performed for sunroofs with two types of buffeting noise control mechanisms. Numerical predictions are compared with the wind tunnel measurements. It is shown that CFD analysis has great potential for sunroof design and development

INTRODUCTION

Wind rush and wind buffeting noise are among the dominant automotive noise sources. Sunroof buffeting is a flow excited resonance phenomenon that usually occurs at low speeds (30-40mph)1. The coupling of the acoustic frequency of the passenger compartment to the periodic instability in the shear layer off the vehicle roof results in the production of high amplitude sound levels. It involves break up of the shear layer into discrete vortices, convection of the vortices with the flow, interaction of the vortices with the downstream edge of sunroof opening and feedback of acoustic disturbance through the vehicle compartment2,3. This inherently unsteady flow phenomenon exhibits complex flow physics and poses severalchallenges for numerical simulations.

Two of the important geometric parameters for the sunroof-buffeting phenomenon are the volume of the passenger compartment and the length of the sunroof opening. Other geometric parameters include the roof curvature and windshield slope. The significance of these

parameters is not yet well known. There are many analytical expressions available to estimate the vortex shedding frequency of the shear layer. One such formula1 for turbulent boundary layers is given by:

F = 1/3(N-1/4) U/L, N=1,2,3,…..

Where U is the free stream velocity, L is the length of the sunroof opening, N is the number of vortices present across the opening. N=2 is generally observed for a typical sunroof on an automobile. No such formula exists for estimation of the noise magnitude in the passenger compartment.

CAE tools are playing an increasingly significant role in the design and development of new automotive vehicles. Building a hardware model is a relatively time consuming process. CAE analysis can be used to identify and improve the performance attributes at a much earlier stage in the design process. CFD analysis for wind noise prediction is one such area that has potential to contribute towards significant reduction in the hardware testing and thus reducing the design time for a vehicle.

The primary outputs required from the CFD analysis are the buffeting frequency and the peak noise magnitude at the buffeting frequency. PAM-FLOW software has been used for sunroof buffeting analysis at GM for the past few years. Numerical predictions are in very good agreement with the wind-tunnel measurements4. This paper presents the contribution of numerical analyses to the design of sunroof-buffeting noise control mechanisms.

There are many methods for suppressing the sunroof buffeting. Two of the methods used in the automotive sunroof industry are:

8th AIAA/CEAS Aeroacoustics Conference & Exhibit<br> <font color="green">Fire17-19 June 2002, Breckenridge, Colorado

AIAA 2002-2550

Copyright © 2002 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.

2American Institute of Aeronautics and Astronautics

1. Deflector at the front of the sunroof opening: The wind deflectors on the front edge of the sunroof opening have been found to be very effective in suppressing the resonance. They deflect the flow such that the vortices pass over the rear edge of the opening. The interaction of the vortex with the downstream edge of the sunroof is thus weaker and therefore less noisy. Introducing notches in the deflector can reduce noise even further. The notches prevent the formation of spanwise-correlated vorticies in the shear layer and thus reduce their strength. The deflectors with notches create more noise in the higher frequency bands, but such penalty is often acceptable because it is less annoying than the low frequency throbbing of sunroof buffeting.

2. Glass comfort position: Another fix to reduce the buffeting noise is the use of so-called comfort positions for the sunroof. In this case, the retracted sunroof glass covers part of the opening. It has been observed that the vorticies grows in strength as they convect downstream. By keeping the sunroof partially closed, vortices interact with the sunroof glass before reaching the rear edge of the opening.

Analysis of both of these methods is presented in this paper. The numerical predictions of the noise magnitude and buffeting frequency from PAM-FLOW are compared with the experimental measurements.

COMPUTATIONAL MODEL

The computational model consists of car in a rectangular box to mimic the experimental setup of noise measurements in a wind tunnel. A full car model with a realistic representation of the interior passenger compartment is used. However, accurate simulation of acoustics properties of the car interior is not attempted. All surfaces in the model and tunnel walls are acoustically rigid. Complex details of the interior like instrument panel, steering wheel are ignored. Figure 1 shows a view of the passenger compartment. Under-hood components are not included in the model. Underbody is approximated using flat panels to reduce the model complexity. All surfaces are patched by a triangular element unstructured mesh. The model is assembled in the pre-processor of the PAM-FLOW package.

Figure 1: Interior view of the passenger compartment.

A good quality mesh is a prerequisite for any accurate numerical simulation. Unstructured tetrahedral mesh for the model is generated using the PAM-GEN, a grid generating software of the PAM-FLOW package. Accurate resolution of shear layer shedding off the roof edge is important to simulate sunroof buffeting. Six layers of boundary layers are used on all the surfaces. The thickness of the first layer is about 2mm. The mesh size on the surfaces near the sunroof opening is 10mm. Surface element size on other parts of the car range from 15mm to 30 mm. The mesh size in the region near sunroof opening is controlled using the volume source lines. Figure 2 shows a view of surface mesh near the sunroof opening. The total model size is about 6 million elements.

The CFD simulations are performed using the software PAM-FLOW. PAM-FLOW is a 3-dimensional simulation code for solving Navier-Stokes equations. The spatial discretization is accomplished via finite element techniques on unstructured grids. Galerkin weighted residual method on linear elements is used. For time discretization, implicit pressure projection scheme is used for incompressible flow solutions. However, the convection term is computed explicitly using the Roe flux splitting scheme with fourth order diffusion. The Laplacian operators obtained for the pressure and velocity correction step are solved using a Preconditioned Conjugate Gradient (PCG) algorithm. The time step is computed from the Courant-Friedrich-Levy (CFL) criteria based on the velocity since the sound speed is infinite in the incompressible flow. PAM-FLOW is a fully

3American Institute of Aeronautics and Astronautics

vectorized and parallelized and can be run in shared (SMP) or distributed (DMP) mode.

Figure 2: Surface mesh near the sunroof opening.

RESULTS AND DISCUSSION

This section describes the results of the CFD simulation of sunroof buffeting for a typical sunroof and for two different types of noise control mechanisms. Important characteristics of sunroof buffeting are described by examining the typical pressure time-history at the driver location. The pressure time-history is used to determine the frequency spectrum of the noise by performing Fourier transformation.

Computational Parameters

Computational parameters are obtained by extensive study5 to ensure mesh-independent solution. Incompressible, transient analysis is performed at 50kph speed. Fixed velocity boundary conditions are enforced on the car surface. Tangential velocity boundary conditions are used for the wind tunnel walls. A boundary layer is not modeled on the wind tunnel walls, as it is not expected to be critical for sunroof buffeting phenomenon. A far-field condition with a secondary condition of static pressure is applied on the wind tunnel exit. A SGS turbulence model with default parameters is used in the analyses.

The MPI version of PAM-FLOW is used to reduce the computation time for analysis.

Calculations are performed using 16-processors on a SGI-3000 machine. The computation time for each case is about 75 hrs.

The fluid domain is initialized to the free stream conditions and the computations are started with a fixed CFL number of 0.5. Initial transients are damped in about 0.5 sec of simulation. Thereafter, the simulation is carried out with a constant time step for acoustic post-processing. The noise spectrum at the driver’s ear is obtained by Fourier transformation of the static pressure time-history at this location.

Typical Pressure Time-history and Frequency Spectrum of Sunroof-buffeting

Unsteady pressure at the driver’s ear location determines the sunroof buffeting noise. Figure 3 shows a part of the typical pressure time-history at 50 kph speed. The time-history shows a dominance of a fundamental mode coinciding with the resonance frequency. Each cycle in the pressure time history corresponds to a passage of vortex from the upstream edge of the sunroof opening to the downstream edge. The minimum pressure in the passenger cabin occurs when the vortex interacts with the rear edge of the sunroof.

-75

-73

-71

-69

3.0 3.2 3.4 3.6 3.8Time (sec)

Pre

ssu

re (

Pa)

Figure 3: Typical sunroof buffeting pressure time-history at the driver ear location.

Figure 4a shows the frequency spectrum of sunroof buffeting noise predicted by the CFD analysis for a typical sunroof on a car at 50kph. The length of the sunroof opening for this case was 0.428m. The estimate for the buffeting frequency using the formula shown earlier is 21.3 Hz. It can be seen from the plot that the CFD prediction agrees well with the analytical results. The spectrum also shows peaks at higher frequencies corresponding to the higher modes of passenger compartment.

4American Institute of Aeronautics and Astronautics

20

40

60

80

100

0 20 40 60 80 100

Frequency (Hz)

SP

L (

dB

)

Figure 4a: Frequency spectrum of sunroof buffeting noise from the CFD analysis.

Figure 4b shows the frequency spectrum of sunroof buffeting noise measured in the wind tunnel for the same car as in the Figure 4a. It can be seen from the plot that the first mode is quite clear in the spectrum but the higher modes are well damped. The magnitude of the first mode for CFD prediction is in excellent agreement with the wind tunnel measurements. Discrepancies in the higher frequency range can be partly attributed to the differences in the hardware model in the wind tunnel and the computational model used in the analysis. The interior of the car in the CFD model does not simulate the sound absorption properties of the real car, which could cause damping of higher frequencies in the wind tunnel measurements.

20

40

60

80

100

0 20 40 60 80 100

Frequency (Hz)

SP

L (

dB

)

Figure 4b: Frequency spectrum of sunroof buffeting noise from the wind tunnel measurement.

Example 1: Design of Wind Deflector for Noise Control

In this example, CFD analysis is used to design a wind deflector for sunroof buffeting noise suppression. Figure 5 shows a view of three sunroofs cases. The baseline design showed

significant sunroof buffeting noise. CFD analysis was used in conjunction with wind tunnel tests to design the deflectors. Two deflector designs were considered. In the first design, a straight edge deflector, referred as Deflector-A, of 30mm thickness was used to deflect the shear layer to prevent the vortices hitting the rear edge of the sunroof opening. In the second design, 10 notches were introduced in the Deflector-A. This design is referred in this paper as Deflector-B.

Figure 5: View of car roof wind deflector designs, (top: Baseline, middle: Deflector-A, bottom: Deflector-B).

Figure 6 shows the plot of noise amplitude of the dominant mode in the frequency spectrum. The wind tunnel tests show that the baseline design has noise amplitude of more than 98 dB and therefore corrective measures are required to reduce the acoustic annoyance in the passenger

5American Institute of Aeronautics and Astronautics

compartment. The addition of Deflector-A (straight edge deflector) reduces the noise magnitude by about 5dB. Using Deflector-B reduces this magnitude even further by about 2dB.

Comparison of the CFD predictions to the wind tunnel measurements shows that the numerical predictions are in good agreement with the experimental results. The computed results show a consistent under prediction of noise amplitude. However, the trend of noise increments is predicted quite accurately. This information is invaluable in evaluating various deflector designs without constructing the time-consuming hardware components.

75

85

95

105

Baseline Deflector-A Deflector-B

SP

L (

dB

)

CFD

Experiment

Figure 6: Peak noise amplitude for sunroof buffeting for various designs at 50kph.

Figure 7 shows a comparison of the frequency of the peak noise amplitude for the three cases. The plot shows that CFD predictions differ from the wind tunnel measurements by less than 1Hz. It is also interesting to note from the figure that Deflector-A results in the peak magnitude at slightly higher frequency while the Deflector-B has resonance at a lower frequency. The CFD predictions also capture this trend excellently.

20

21

22

23

24

25

Baseline Deflector-A Deflector-B

Fre

qu

ency

(H

z)

CFD

Experiment

Figure 7: Frequency of the peak noise amplitude for sunroof buffeting for various designs at 50kph.

Example 2: Comfort Position for Noise Control

This section describes an application of CFD methods to evaluate the effects of comfort stop position on sunroof buffeting. Figure 8 shows a view of sunroofs for the two cases. The picture with the sunroof glass shows the glass in the fully open position. In the fully open position, sunroof glass partly covers the opening, thus reducing the opening length from 0.57m to 0.34m. The reduced opening length allows the vortices to interact with the glass before reaching the rear edge of the opening. .

Figure 8: View of the sunroof opening with and without the sunroof glass.

Figure 9 shows the plot of noise magnitude of the dominant mode in the frequency spectrum for the two cases. The CFD predicted magnitude is smaller than the wind tunnel measurements, just like the first example. The use of retracted glass as a comfort stop helps in totally eliminating the buffeting phenomenon. The noise magnitude of 80dB or less in the frequency range showed here is acoustically insignificant.

6American Institute of Aeronautics and Astronautics

40

60

80

100

120

w/o Sunroof glass w Sunroof glass

SP

L (

dB

)CFD

Experiment

Figure 9: Peak noise amplitude for sunroof buffeting for various designs at 50kph.

CONCLUSIONS

PAM-FLOW CFD software is used to simulate the sunroof-buffeting phenomenon. The characteristics of the pressure time-history and frequency spectrum of the buffeting noise are described. Noise is computed for two sunroofs designs with noise control mechanisms. In the first case, a wind deflector is used to reduce the buffeting noise while in the second case a comfort position stop is used. Both cases compared well against the wind tunnel measurements. The results show that CFD can be used to aid the acoustic design of sunroofs for automotive vehicles. Furthermore, the influence of various geometric parameters of a vehicle can be studied to understand the phenomenon and avoid sunroof buffeting altogether.

ACKNOWLEDGEMENT

The authors would like to thank James C. Zunich and David Horgan from the Noise and Vibration Center at General Motors for providing the data of wind tunnel experiments.

REFERENCES

1. Ota, D.K., Chakravarthy, S.R., Becker, T. and Sturzenegger, T., ``Computational Study of Resonance Suppression of Open Sunroofs’’, Journal of Fluids Engineering, Vol 116, pp 877-882, Dec. 1994.

2. Nelson, P.A., Halliwell, N.A., and Doak, P.E., ``Fluid Dynamics of Flow Excited Resonance, Part I: Experiment’’, Journal of Sound and Vibration, Vol 78(1), pp 15-38, 1981.

3. Nelson, P.A., Halliwell, N.A., and Doak, P.E., ``Fluid Dynamics of Flow Excited Resonance, Part II: Flow Acoustic Interaction’’, Journal of Sound and Vibration, Vol 91(3), pp 375-402, 1983.

4. Karbon, K. and Kumarasamy, S., ``Computational Aeroacoustics Applications in Automotive Design’’, First MIT Conference on Computational Solid and Fluid Mechanics, MIT, June 2001.

5. Singh, R, ``Sunroof Buffeting Simulations Using PAM-FLOW’’, Ameripam-2001, PAM-FLOW User’s Conference, Detroit, Nov. 2001.

6. PAM-FLOW 2000, User’s manual. ESI Group, www.esi-group.com, 2001.