1. Introduction
When buffeting occurs in a passenger vehicle, strong pressure fluctuations inside the cabin are induced by the external flow over an opening of the cabin(sunroof or side-window) creating an unpleasant discomfort to the passengers. The coupling happening between the external flow over the opening and the acoustic behavior of the cabin is well understood and can be measured in a wind-tunnel. But car makers need accurate means of simulating buffeting to predict its occurrence and prevent it during the development of vehicles.
During buffeting, the strength of the sound-pressure level(SPL) inside the cabin will depend on several factors such as the size of the opening, its location, the external flow speed and the characteristics of the cabin(leakage, wall stiffness, impedance, etc.). Buffeting has been studied and simulated for some years now and methodologies have been developed to include some of the real world effects(RWE) that influence the SPL during buffeting(1,2).
So on a real car, many factors will influence the results. To stress out the ability of a numerical method to reproduce buffeting, an experiment can be done where many parameters are controlled to minimize their impact. This was done in the past by a group of German car makers(3,4) and this is the approach that was adopted by HMC for this work.
2. Experiment and Simulation
2.1 Experimental Stup
The physical model used for testing was a simple wedge shaped box with and opening on the roof referred to as the Hyundai simple model (HSM). The HSM was made of thick aluminum walls mounted on a metal frame. The inside of the model was padded with absorbing material to change the acoustic response of the cavity. The model was sealed to avoid any leakage flow. The HSM is shown in Fig. 1. The Q-factor of the cavity was evaluated at 10.3 using an acoustic response test(ART).
Fig. 1Configurations of the HSM test in the Hyundai wind-tunnel
The model was made to sit on the floor of the wind tunnel. The time history of the pressure inside the model was recorded by a microphone and each pressure time signal was post-treated to get the spectrum. The maximum point of the spectrum and its frequency were extracted for flow velocities ranging between 20 km/h and 100 km/h. Experimental results are presented on Fig. 3 and 4 below.
A second measurement was done at 60 km/h on the roof of the same model with the roof opening closed. The velocity profile was measured at 3 different locations as shown on Fig. 2.
Fig. 2Positions on the roof where the boundary layer profiles were measured
2.2 Buffeting
When an external air stream flows over the opening of a cavity a shear layer is formed. A shear layer is an unstable flow structure that has the tendency to roll up into vortices. These vortices are convected downstream over the opening at a speed slightly below free stream velocity and interact with the trailing edge of the opening. This fluid-structure interaction generates a pressure wave that propagates inside the cavity where it is reflected by the walls. This acoustic wave then comes back to the opening to interact with the shear layer(5).
Any cavity with a given opening behaves like a Helmoltz resonator with a given natural frequency. The inside volume of the cavity acts as a spring attached to an equivalent air mass associated with the cavity’s opening. The vortex shedding frequency in the shear layer increases linearly with free stream velocity. When the shedding frequency gets closer to the natural frequency of the cavity, a lock-on phenomenon occurs where the acoustic feedback from the cavity initiates the formation of the next vortex in the shear layer. This fluid-acoustic coupling is characterized by an increase of the pressure amplitude in the cavity that is called “buffeting”.
As the external velocity is further increased, the shedding frequency becomes larger than the natural frequency of the cavity and the coupling becomes weaker, leading to a decrease of the pressure amplitude inside the cavity. This is the offset phase.
At higher velocities, the shear layer brakes down in smaller vortices that are still interacting with the trailing edge of the opening but over a broader range of frequencies. The cavity is then acoustically excited over a wider range of frequencies but with smaller amplitudes. The sound pressure level is then lower than when buffeting occurs, but it increases with external velocity since the mean momentum of the vortices in the shear layer is proportional to the external velocity.
2.3 Setup and Simulation Procedure
The simulations were performed using PowerFLOW version 4.3b. PowerFLOW is a Lattice Boltzmann method(LBM) based CFD software using a VLES turbulence model. It is a very efficient transient flow solver used by many OEMs(original equipment manufacturer) in the development of car worldwide. The background of the method and the details of its implementation can be found in the references(6~16).
The basic setup was based on the sunroof buffeting best practices recommended by Exa, which was developed and validated with different cars against wind-tunnel measurement. A variable resolution scheme with a minimum cell size of 1.5 mm was used in these simulations. Simulations were done with a symmetry plane and pressure was recorded in the middle of the cavity. Real world effects obtained from the acoustic response test(ART) were implemented in the simulation model to match the acoustic response characteristics of the cabin between test and simulation.
2.4 Simulation Results
Simulations were done for velocities ranging from 20 km/h to 90 km/h.
Simulation results are shown in Fig. 3 and 4 for SPL and frequency respectively. The SPL curves show good correlation between test and simulation for the critical velocity range where buffeting occurs. The maximum buffeting occurs at 50 km/h, as in the experiment. SPL for 30 km/h and 60 km/h are at the same level as the experiment. The 40 km/h and 50 km/h SPL are only 4 dB and 2.5 dB lower than experimental values respectively. SPL for velocities above 60 km/h are over predicted by about 6 dB.
Fig. 3SPL results with the full Q-factor
The frequency curves show good correlation between test and simulation results. As can be seen on Fig. 4, the buffeting frequency at 60 km/h is nearly on the experimental value. The worst deviation is by 1.5 Hz at 40 km/h and 80 km/h.
Fig. 4Frequency results with the full Q-factor
Figure 5 shows the pressure signals inside the cavity obtained for different free stream velocities. At the onset of buffeting(30 km/h), the amplitude is low with signs of intermittency. At 50 km/h a strong buffeting is present with a very regular frequency and little modulation of the pressure amplitude. Increasing the free stream velocity to 60 km/h brings the system into the offset mode where strong intermittency shows up again. At 90 km/h the pressure signal becomes more chaotic with small fluctuations immediately followed by strong ones.
Fig. 5Simulated pressure time signals at different velocities shown against the SPL curves
So the largest discrepancy exists at higher velocities( 70 km/h and above). In this velocity range, the feedback mechanism is not the same as in the buffeting range(40 km/h to 60 km/h). At higher velocities, the acoustic feedback is much less important. The cavity is purely excited by the shear flow interacting with trailing edge of the sunroof opening, and this excitation depends on the structure of the shear flow. Obviously the shear flow obtained in the simulation so far is exciting the trailing edge too much and the maximum SPL in the cavity is too high. Important aspects of shear flow simulation are the boundary layer development of the flow that leads to the shear layer and the possibility of the shear layer to breakdown into smaller vortices.
When analyzed more in depth, the flow around the HSM for this study is trickier than what it can seem at first glance. One peculiar flow structure requires more attention and can turn out to improve both the missing SPL at 40 km/h and 50 km/h, and improve the over prediction at higher velocities. This structure is a small flow separation found on the roof right after the slanted front-roof corner. This small flow separation changes the shape of the velocity profile that arrives at the sunroof and influences how the shear layer will develop over the sunroof opening. It was reported by HMC engineers that a flow separation was observed using smoke injection at the roof front corner during the tests in the wind-tunnel.
2.5 Simulation Results – Higher resolution
To better understand the impact of the separation on the shear layer and on the excitation of the cavity, some simulations were done with a higher resolution on the roof and in the shear layer since the size of the flow structures generated in the separation region in front of the sunroof opening are very small. One velocity from the buffeting range(50 km/h) and the offset range(80 km/h) was selected for these higher resolution runs.
At 50 km/h, from the y-plane pictures of vorticity at four different times on Fig. 6, it can be seen that the separation ahead of the opening is intermittent. It shows only on the two last pictures. The four pictures cover a typical cycle of buffeting(about 1/30 sec). In buffeting mode, the shear layer rolls up to create large vortices and the small separation ahead of the opening only creates a small loss of momentum so the vortices interact more strongly with the trailing edge of the opening, and the excitation of the cavity is stronger. The SPL predicted is now exceeding the experimental value by less than 1 dB. If the boundary layer ahead of the opening is not enough resolved, the separation will be over predicted and the momentum loss will be larger. This will lead to a lower excitation of the cavity and a lower SPL. This is what was observed with the lower resolution results presented before.
Fig. 6Y-plane vortices magnitude at 50 km/h with high resolution on front edge
At 80 km/h, the Reynolds number is higher and the separation region more important. Many small vortices are produced ahead of the opening and convected towards the shear layer. At this velocity the feedback mechanism is not as important as before and the shear layer is composed of many vortices of different sizes. A certain periodic clustering of vortices can be observed in the pictures of Fig. 7 also evenly spaced over a period of 1/30 sec. This shows up as a bump in the pressure spectrum inside the cavity. This is due to a weak acoustic feedback that is still present. But the coherence of the shear flow structures is much less important and the global excitation of the cavity is lower. This is reflected in the lower SPL value.
Fig. 7Y-plane vortices magnitude at 80 km/h with high resolution on front edge
With the higher resolution, PowerFLOW is able to capture this mechanism and the predicted SPL get very close to the one measured(see Fig. 8 and 9). The 80 km/h SPL value is now only over predicted by less than 2 dB. When the flow is not enough resolved, the vortices coming from the front are larger and the excitation of the cavity also. This is why the SPL was over predicted before.
Fig. 8SPL and frequency for high resolution results for 50 km/h
Fig. 9SPL and frequency for high resolution results for 80 km/h
The frequency at 50 km/h gets slightly higher and at 80 km/h predicted frequency just gets bellow the experimental value. Overall it is an improvement over the previous results.
Because PowerFLOW captured the different mechanisms in buffeting mode and in the offset of buffeting, it is expected that using the higher resolution setup for all velocities would reproduce the experimental values very closely. At 90 km/h and 100 km/h, the vortices in the boundary layer in front of the opening will have the tendency to break down in even smaller structures but their mean convective velocity will be higher and SPL should increase slowly as it is seen with the lower resolution results and as it is seen experimentally.
2.6 Boundary Layer Profiles
The HSM was also simulated with the sunroof opening closed to obtain the boundary layer profile at three locations on the roof with a free stream velocity of 60 km/h(16.67 m/s). The comparison to the experimental velocity profiles is shown in Fig. 10.
Fig. 10Comparison of boundary layer profiles(profiles 2 and 3 have been shifted by 5 m/s and 10 m/s for clarity)
The simulated profiles compare well to the experimental measurements at the 3 locations with a slight tendency to underestimate the velocities within the boundary layer. The shapes of the three profiles are identical to the experimental profiles, showing that the flow structures simulated within the boundary layer are reproduced correctly. The underestimation of the velocity within the boundary layer is probably due to a slight over prediction of dissipation in the separation region at the front corner. Ways to improve the results are under investigation. Outside of the boundary layer the comparison to the experimental velocities is very good with errors well within experimental uncertainty.
3. Conclusions
The buffeting behavior of a simple wedge box was tested in the Hyundai wind-tunnel and simulated with PowerFLOW. Peak SPL levels obtained from the simulation correlate well with the test results for the buffeting velocities. Simulation over-predicted the SPL levels for the higher velocities(in the offset region) with the initial setup. By increasing the resolution in the boundary layer in front of the opening, it was demonstrated that results can be improved both in the buffeting range and in the offset of buffeting at higher velocities.
A separation region present just behind the front corner was identified as a key element in determining the SPL over the whole velocity range. At velocities below 60 km/h, the separation is intermittent and does not reduce much momentum close to the surface, leading to high SPL. At velocities above the 60 km/h, the separation is permanent and produces small vortical structures within the boundary layer which results in a much less coherent excitation of the cavity and lower SPL. As observed experimentally.
The velocity profiles of the boundary layer developing over the closed roof configuration compare well to the measurements with a slight tendency to underestimate the velocity within the boundary layer. The velocity prediction outside the boundary layer is very good.
References
- Crouse, B., Balasubramanian, G., Senthooran, S., Freed, D., Ih, K. and Shin, S., 2009, Investigation of Gap Deflector Efficiency for Reduction of Sunroof Buffeting, SAE Paper 2009-01-2233.
- Crouse, B., Senthooran, S., Balasubramanian, G., Freed, D. and Karbon, K., 2007, Computational Aeroacoustics Investigation of Automobile Sunroof Buffeting", SAE Paper 2007-01-2403.
- Islam, M., Decker, F., Hartmann, M., Jäger, A., Lemke, T., Ocker, J., Schwarz, V., Ullrich, F., Schröder, A. and Heider, A., 2008, Investigations of Sunroof Buffeting in an Idealised Generic Vehicle Model - Part 1: Experimental Results, 29th AIAA Aeroacoustics Conference, Paper 2008-2900.
- Islam, M., Decker, F., Hartmann, M., Jager, A., Lemke, T., Ocker, J., Schwarz, V., Ullrich, F., Crouse, B., Balasubramanian, G. and Mendonca, F., 2008, Investigations of Sunroof Buffeting in an Idealised Generic Vehicle Model - Part 2: Numerical Simulations, 29th AIAA Aeroacoustics Conference, Paper 2008-2901.
- Kook, H. and Mongeau, L., 2002, Analysis of the Periodic Pressure Fluctuations Induced by Flow Over a Cavity, Journal of Sound and Vibration, Vol. 251, No. 5, pp. 823-846. https://doi.org/10.1006/jsvi.2001.4013
- Bhatnagar, P., Gross, E. and Krook, M., 1954, A Model for Collision Processes in Gases. I. Small Amplitude Processes in Charged and Neutral One-component System, Physical Review E, Vol. 94, pp. 511-525. https://doi.org/10.1103/PhysRev.94.511
- Bres, G., Perot, F, and Freed, D., 2009, Properties of the Lattice-Boltzmann Method for Acoustics, 30th AIAA Aeroacoustics Conference, Paper 2009-3395.
- Chen, H., 1998, Volumetric Formulation of the Lattice Boltzmann Method for Fluid Dynamics: Basic Concept, Physical Review E, Vol. 58, pp. 3955-3963. https://doi.org/10.1103/PhysRevE.58.3955
- Chen, H., Orszag, S., Staroselsky, I. and Succi, S., 2004, Expanded Analogy between Boltzmann Kinetic Theory of Fluid and Turbulence, Journal of Fluid Mechanics, Vol. 519, pp. 307-314.
- Chen, H., Teixeira, C. and Molvig, K., 1998, Realization of Fluid Boundary Conditions via Discrete Boltzmann Dynamics, International Journal of Modern Physics C, Vol. 9, No. 8, pp. 1281-1292. https://doi.org/10.1142/S0129183198001151
- Chen, S. and Doolen, G., 1998, Lattice Boltzmann Method for Fluid Flows, Annual Review of Fluid Mechanics, Vol. 30, pp. 329-364. https://doi.org/10.1146/annurev.fluid.30.1.329
- Chen, S., Chen, H., Martinez, D. and Matthaeus, W., 1991, Lattice Boltzmann Model for Simulation of Magnetohydrodynamics, Physical Review Letters, Vol. 67, No. 27, pp. 3776-3779. https://doi.org/10.1103/PhysRevLett.67.3776
- Li, Y., Shock, R., Zhang, R. and Chen, H., 2004, Numerical Study of Flow Past an Impulsively Started Cylinder by Lattice Boltzmann Method, Journal of Fluid Mechanics, Vol. 519, pp. 273-300. https://doi.org/10.1017/S0022112004001272
- Qian, Y., d'Humieres, D. and Lallemand, P., 1992, Lattice BGK Models for Navier-Stokes Equation, Europhysics Letters, Vol. 17, No. 6, pp. 479-484. https://doi.org/10.1209/0295-5075/17/6/001
- Rubinstein, R. and Barton, J. M., 1990, Nonlinear Reynolds Stress Models and the Renormalization Group, Physics of Fluids A, Vol. 2, No. 8, pp. 1472-1476. https://doi.org/10.1063/1.857595
- Teixeira, C., 1998, Incorporating Turbulence Models into the Lattice-Boltzmann Method, International Journal of Modern Physics C, Vol. 9, No. 8, pp. 1159-1175. https://doi.org/10.1142/S0129183198001060