This article describes features of a car which is General Motors' (GM) technology demonstration entry in the Partnership for a New Generation of Vehicles, the PNGV program. Compared to the EV1, the ultra-efficient two-passenger electric vehicle GM has been selling in California since 1996, the new concept car has 34 fewer counts (0.034) of aerodynamic drag. The engineers needed to establish the vehicle's architecture early, knowing that any mistakes there would be irreversible. They had to evaluate armrest positions and side-window clearance. By wearing its cooling-air intakes on the rear fenders-a benefit that comes with mounting the engine in back—the new shape borrows from sibling EV1's success with low ram air inlet. While the shape investigation was under way, an EV1 test mule aided the concurrent development of features. A full- size model of the technology demonstration shape was ready for wind-tunnel testing by June 1998.


Everyone did a little airfoil research as a kid. Aeropalmics, the field's formal name, required nothing more than a car moving along at 60 or so, a warm, dry day that invited open windows, and a willingness to stick a hand out into the airstream and let it fly.

Hold your hand out the window of one new car, and you will get 20 fewer miles to the gallon. Close the window, and it will be so quiet you'll think you are in a library. You won't be checking your teeth in the side mirror though, for this car has none-too much drag.

You can't buy one yet. Premiering in January at the North American International Auto Show, the car is General Motors' technology demonstration entry in the Partnership for a New Generation of Vehicles, the PNGV program. With a coefficient of drag, or CD' of 0.163, the vehicle represents more than a halving of the drag associated with a typical midsize sedan. Compared to the EV1, the ultra-efficient two-passenger electric vehicle GM has been selling in California since 1996, the new concept car has 34 fewer counts (0.034) of aerodynamic drag.

The record-low drag coefficient partly fulfills a GM program for meeting the New Generation of Vehicles goal of a car that costs the same to own and drive as today's average family sedan but gets three times the gas mileage. The car must perform as well as a Taurus, a Lumina, or a Concorde, be every bit as safe and pollution-free, and be 80 percent recyclable. To get there, GM has set off on a course to double vehicle efficiency and increase propulsive efficiency by half. Production prototypes representing the New Generation of Vehicles are due from GM, DaimlerChrysler, and Ford in 2004.


Rear Engine Raindrop

Although the goal of tripling the fuel economy for a midsize sedan was set in September 1993, actual shape studies did not begin at GM until November 1997.

In the intervening years, the three automakers and the government debated just how they could make their novel partnership work. By late 1997, however, a six-person team comprising engineers, aerodynamics and wind tunnel experts, and industrial designers had assembled at the GM Center for Advanced Technology Vehicles in Troy, Mich. Ron York, GM's PNGV director, challenged the group to develop an aerodynan1ic sedan, and gave them a finish line a mere eight, or possibly 10, months away.

Some members of the team had worked on the EV1. The experience of building that vehicle provided the group with an entry point for starting the new shape. If the team could match the drag coefficient of the EV1 (0.19), then the New Generation car could probably reach its 80 mpg goal, especially considering the many other technological seedlings outside of aerodynamics that were expected to take root.

The engineers needed to establish the vehicle's architecture early, knowing that any mistakes there would be irreversible.

They ran consumption and loss scenarios for typical vehicles, according to vehicle integration engineer George Claypole, based upon what had been learned in developing the EV1. That car, with a frontal area of 0.188 square meter s, achieved a CoA (the product of drag coefficient and frontal area) of 0.366. The engineers guessed that the frontal area for the new shape would come in at 2.05 square meters. To keep drag down, they would incorporate many of the features that made the EV1 so sleek: a flat under body, closed-down wheel housings and openings, and low ram air inlet for cooling air.

Speaking of cooling, a turbo charged diesel engine would power the new car, adding appreciably to a thermal burden that had been limited mostly to cooling the occupants in the electric-powered EV1. Passing cooling air over an engine radiator was a problem the EV1 engineers had not had to solve. But, from the beginning, the new project had embraced the idea that styling, passenger accommodations, and subsystems compliance would progress in parallel with the development of the car's outer wrapping. That strategy began to pay off early.

The ideal aerodynamic shape is a raindrop, of course. Yet, forcing that shape upon a functional car body posed a bit of a problem. Squashing the nose made very little space available to an engine, and stretching the tail added a fair amount of weight in unused structure. A rear-mounted engine could relieve a hot of concerns. Because it could be mounted transversely between the two nonsteerable rear wheels, an engine in the rear would allow the frontal area to shrink. Frontal area would then depend on the needs of the passengers alone, not the demands of the drivetrain.

As another benefit, a rear-mounted engine would eliminate an exhaust system that spanned the car's entire length. No long, heavy pipe, catalytic converter, or muffler would dangle in the airstream. Thus, the new sedan could retain the flat underbody of the EV1 along with its minimal ground clearance.

Perhaps the biggest benefit of an engine in the rear, though, would be how it could eliminate front air intakes and underbody discharges. Such major surface features cost dearly on the Co ledger. Instead, the engine heat exchanger would take its air as needed through inlets along the rear fender surfaces. Exhausting cooling air out the back would actually help efficiency by filling the wake.

With an architectural concept outlined, the team was now a month closer to the finish. But it had yet to generate any three-dimensional designs. First, according to Mike Kutcher, group manager for exterior work, "They had to define the hard points."

Hard points were the minimal outside vehicle boundaries that would allow sufficient room. for safety systems, passengers, and wheel tracking. In keeping with the goal of developing a car as safe as anything currently available, the new vehicle's safety features would include crush rails and bumpers. The engineers also had to decide how much clearance to allow for the car's birdcage, that is, the automobile's major structure.

As for the passenger compartment, the engineers had to determine how much space to leave above the heads of drivers and riders. They had to evaluate armrest positions and side-window clearance.

For wheel track boundaries, the engineers asked how much space the steerable front wheels would need. How much room would be needed for tire flop, the dynamic envelope that inscribes the wheel's full range of motion? And how far inboard from the front wheels could the rear wheels be? The answer to this last question , as it had on the EV1, would tell the PNGV team how close they could sneak up on the raindrop ideal.

The hard points thus defined, it was time to shrink-wrap a skin on the concept and quantity the frontal area. Next, the engineers and aerodynamicists, using GM's proprietary artificial intelligence-based neural network and CFD software, expanded this shape beyond a taut-skin minimum, rounding it out to develop sleeker contours. "The software was used as a diagnostic tool to see how they could change surfaces and understand the consequences of those changes on the aerodynamic drag," said V. Sumantran, lab group manager at the GM Vehicle Analysis and Dynamics Laboratory. Using rapid prototypes made in one-eighth and one-tenth scale on a laminated object modeler system from Helisys Inc. of Torrance, Calif., designers could go from mathematical analysis on a computer to empirical verification in the wind tunnel in as few as five days.

Wind tunnel testing proved more efficient than CFD analysis for checking drag at various yaw angles. Research had indicated that cars normally experience headwinds averaging 7 mph and acting at an angle of 8 degrees from head on. Since the PNGV vehicle would go beyond its laboratory beginnings and roam the streets one day, the engineers asked often how it would fare on the open road. In the wind tunnel, the models faced the airstream nose first and shoulder first, too; they were swept through yaw angles as high as 12 degrees from zero.


Designers Touch

At this point, with the engineers and aerodynamicists having spent quite a number of nights in the wind tunnel, the car was within 10 to 20 percent of its aerodynamic goal, said Kutcher. They had converged on a mainstream shape. Now the designers could contribute to the effort by sculpting from the idealized airfoil shape an auto body that had less in common with an "aerodynamic potato."

From the start, the designers "wanted the exterior to walk hand in hand with the car's mechanical underpinnings," said Mike Pevovar. H e and co-designer Brian Smith saw right away that the roof of the passenger compartment was too low for easy entry and egress. They wanted to bump up roofline height, and increase "tumble home," the slope the side windows follow as they drop away from the roof. (A truck, for instance, has little tumble home.)

No strangers to aerodynamics, Pevovar and Smith knew the changes they were proposing would add to the frontal area, with a resulting rise in the CDA value. Also, they wanted to use planar surfaces that intersected at angles and creases-again , good for style but thought to be bad for efficient airflow.

With these changes in mind, the two designers met with the engineers and aerodynamicists, beginning a series of discussions that would come to be known as the "120 dB conversations." The engineers' heavily researched shape was fairly efficient aerodynamically and close to the target values for drag; to change it as the designers were now proposing was surely blasphemous. After all, hadn't the CFD analysis been validated in the wind tunnel?

Eventually, the designers cajoled the engineers into agreeing to a variant. The designers would form a second clay model in one-third scale while the engineers built and refined their own shape at one-third size. Model development thus diverged onto two parallel paths. The designers' version would differ from the research version in two respects, however: It would include styling cues that had been left off the first model, and, where possible, it would use planar surfaces aligned with streamed surfaces that had sharp, rather than rounded transitions.

By March 1998, th e engineers and the designers had committed their ideas and visions to one-third-scale models of clay. A friendly bake-off was about to begin. The two teams were given four hours apiece to tune their respective shapes in the wind tunnel. Both models needed wind tunnel tuning because there were many details that simply could not be evaluated effectively, either mathematically or through one-eighth-scale testing. Four hours in the tunnel would give the teams time to remove or add a little clay here and there.

The research model was the first one in. Starting with a CD of 0.20, the engineers were able to reduce that value to 0.19 within four hours. Then the designers put their variant in the wind tunnel. To everyone's surprise, the stylized model returned a CD of 0. 18 on its first run.

Needless to say, the designers greeted their result first with disbelief, then with joy. "We had our credibility on the line," Pevovar recalled . Whether the outcome was a result of technical inspiration, artistic intuition, or just plain happenstance no one is saying, but the shape that would ultimately be modeled full-size would be bringing along with it a lot of styling points. After their initial shock dissipated, the teams regrouped, and by session 's end had succeeding in reducing the model's drag coefficient to a slippery 0.46.

The goal of a CD equal to 0.1 9, set at the beginning of the program, budgeted 30 counts for features drag. Elements such as wipers, wheel covers, mirrors, door handles, and cut lines all could affect how the car would slice through the air. With the basic shape now defined, the group could finish up its work on features, many of which had been undergoing evaluation all along.

Side mirro rs were one group of features given early attention. Adding 15 counts of drag by themselves, the mirrors simply could not remain if the car was really going to cheat the wind out of 80 mpg. Small cameras would mount in their place; facing rearward, they would send a screen image to the driver.

While the shape investigation was under way, an EV1 test mule aided the concurrent development of features. Using this full -size test stand, engineers evaluated feature drag independently of the shape tests without having to wait for the full-size four- door model to emerge. In their preliminary assessments of the path through which cooling air would enter and exit the heat-exchanger loop, for example, engineers evaluated square and rectangular inlets and peripheral, side- edge, and core outlets using CFD. They verified their findings on the EV1 test mule.

Ground clearance was also investigated in parallel with body shaping. Ground clearance normally varies according to the number of riders inside a car. But a low-riding automobile is an efficient one. This reasoning was one of several arguments favoring a rear- engine location. An exhaust system running back from the front of the car would need extra space to fit between the underbody and the road. Ground clearance, in fact, would have so great an effect on the car's mileage that the GM engineers decided to use active control. Air springs and a ride- height sensor would compensate for height changes as vehicle loading varied.

Other mechanical features included unvented wheel discs to avoid the rotational drag associated with cooling the brake rotors. Regenerative braking, another feature, would help in slowing the vehicle without converting the energy of motion into heat. And batteries would be stored beneath the seats as a way of reducing frontal area.

Despite a wholesome budget of 0.030 given the team for features drag, it came in well below that mark. Even taking into account a rocker fence along the rear trunk edge, another styling concession added by the designers, the features added a mere 0.007 to the CD'


Scaling Up

A full- size model of the technology demonstration shape was ready for wind-tunnel testing by June 1998. In its initial trial, the model came in with a CD of 0.151. As group members began to add features and refine the shape to suit everyone's tastes, they discovered that the rearview cameras that replaced the mirrors added no drag at all. The same was true of the windshield wipers.

By August, the group had arrived at a final form that pleased the designers, the engineers, and the aerodynamicists. Its last run in the wind tunnel returned a CD of 0.1 63. In 10 months, the group had set a record-low drag coefficient for a midsize sedan. It had surpassed the aerodynamic drag of the two-person EV1 and had even fashioned an aerodynamic auto that managed to avoid any comparison with potatoes.

But is there a lesson here that somehow contradicts the pilot's adage of "trust your instruments"? How is it that the highly researched vehicle came in having more drag than the stylized version?

Asked this question, designer Pevovar said, "From the beginning, the engineers were seeking the perfect aerodynamic shape." The designers took a slightly less stringent, more intuitive approach. They sought ways to fit the shape to the occupants.

By raising the roof, though that meant increasing frontal area, the designers enlarged passenger space at the same time that they smoothed transitions among the hood, the windshield, the roof, and the tail. The net effect was less drag and n1.ore passenger space. Since they were never seeking aerodynamic perfection, the designers could drift away from the tightly constrained shape to which the engineers had been anchored.

Vehicle integration engineer Claypol e added another point. The deadline of eight to 10 months, set by York in November, was coming around fast by the time of the bake-off in March. Much work remained, including developing the full-size model and finishing the features. The two clay models actually came in quite close to each other, Claypole said.


If they had the luxury of time, the team members might have been able to refine the research shape below the value that the stylized shape had reached. But they didn't have that luxury. Given a model that did better in its first trial than the other had done after four hours of testing, the group could only invest what time remained into the model promising the greatest return.

Consider, too, that many surface details were added for their aesthetic value; the car, after all, would have to sell. Once computational methods began consistently yielding result within 10 to 20 percent of the target, an almost unlimited number of surface features could be applied to bump up the vehicle's appeal. It so happened that the particular features chosen also managed to reduce drag.

Mike Kucher puts it this way: " For any given set of packaging hard points and basic shapes, there will be an infinite number of final surface solutions to achieve a desired aero performance. Once you have identified the best basic-shape proportions, the final surface details can be optimized to achieve the best end result."

Eighty mpg or not, though, there will probably be a kid riding around in a car with windows down and hand flying. A researcher's work never ends.