Page 53 - Computational Fluid Dynamics for Engineers

38 1. Introduction

Rear flap
Canard wing Rear under flap
Chin spoiler

0.5 f 1 0.4f
J 0.4 loi /
«
| 0.3
S
^ 0 . 2 -0.4 -0.2 +>* 0.2 0.4
O jr CI (experiment)
0.1! . / / -0.2
/ 1
y
0.0 0.1 0.2 0.3 0.4 0.5 . * -0.4
Cd (experiment)
DRAG LIFT
Fig. 1.36. Comparison of computed (DNS) and measured (symbols) drag and lift coeffi-
cients for a sports car with various aerodynamic devices [29].

In summary, if one considers that the requirement for accuracy in drag pre-
diction of cars is within 0.5 percent and that the best CFD can produce is
within 5 percent, it may appear that CFD has a long way to go in achieving
this goal. However, the picture is not as bad as it appears because the other
aspects of car aerodynamics do not have to be predicted as closely as drag,
which is one of the main factors contributing to fuel consumption and amounts
to official 46 percent of the total for a midsize U.S. car design on a highway.
The computing times associated with the obtainable accuracy are still excessive
in comparison with times required for wind-tunnel testing. It should be pointed
out that the previously cited computing time is only to obtain one data point.
Looking ahead, it is likely that these computer times will be reduced drastically
as massively parallel computers become available. It is likely that the problems
with long CPU-time and the limitations in the number of panels will eventually
be reduced to a nuisance level. It is also clear that with more computer power
becoming available, there is going to be more emphasis placed on improving
physical modeling of turbulence, since understanding this process is essential to
obtaining more accurate results.