Experimental and computational investigation of the flow field around 2- and 3-dimensional automotive shapes

One of the major objectives of automotive aerodynamics research is the reduction of vehicle drag. In addition to the traditional force balance measurements, a detailed investigation of the flow field is important for further reduction of drag in the future cars and for better airflow management within and around them.

In the present work, the flow fields around 2- and 3-D Mercury Cougar and Ford Thunderbird scale models were investigated in a wind tunnel flow. Smoke visualization was employed for qualitative investigation. Center plane velocities and turbulence intensities near the surface and wake regions of the models were measured by Laser Doppler Velocimetry (LDV). Surface pressures and wind tunnel floor pressures were also measured at the center plane. An experimental data base for the measured quantities was thus established which will be useful to qualify the present and future computational fluid dynamics (CFD) codes. Two CFD codes: FORDC-2 and FLUENT were used to simulate and compute the flow quantities for the flow around the 2-D models. The computed results are tabulated and compared graphically with the experimental results. The flow fields around the 3-D models are also compared with those around the 2-D models.

The visualization studies showed a marked difference in the separation characteristics of flow around the 2- and 3-D models. The LDV technique is very effective in the determination of velocities and turbulence levels. In the model wake, the data rate, however, was very low. The uncertainty in the measured quantities is high compared to those in regions prior to separation. Poor signal to noise ratio of the LDV signal in the wake region is the cause for large measurement uncertainties. The measured turbulence intensities were higher than expected at several locations due the inherent phase noise in the LDV signal. The CFD methods predicted the flow parameters reasonably well up to the separation region. There were significant deviations between the predicted and experimental values of turbulence intensities in the wake region. In addition, the predicted surface pressures and the near-surface velocities were in disagreement with the experimental values. This is attributed to the inadequacy of the "near-wall" model and the K — e turbulence model incorporated in the two codes that were tested.